Welding the Iraq Petroleum Company pipeline across the Plain of Esdraelon, July 1933. Photograph: American Colony (Jerusalem), Photo Department. Library of Congress, Prints and Photographs Division, LC-DIG-matpc-15775.

At five in the afternoon on 2 November 1956, the Syrian government cut the Iraq Petroleum Company’s communications, according to the American ambassador. The company stopped pumping early the next morning. Iraq’s oil had lost its passage to the Mediterranean.^{⁠[⁠10⁠]}

That afternoon, three Syrian ministers met the ambassador, James Moose. The prime minister was there, along with the acting foreign minister and the minister of public works. They believed a pumping station had been seriously damaged. They said they did not know where, or who had done it.

“Visibly agitated,” Moose wrote of the men before him. He reminded them that the prime minister and acting foreign minister had already assured him the installations would be protected. All three now promised every possible assistance with repairs. The public works minister, himself an engineer, offered to help personally.^{⁠[⁠10⁠,⁠25⁠]}

The CIA’s bulletin of 12 November described two pumping stations destroyed and a third extensively damaged. Yet the company’s local superintendent believed some oil could move before they were rebuilt. New pipe could bypass the wreckage, allowing the surviving station in Iraq to push about 80,000 barrels a day towards the coast.

“This flow could be established in a few days if the Syrian government were willing.”^{⁠[⁠11⁠]}

Who, in Damascus, could give that permission?

· · ·

Long before the sabotage, British ministers had worried about France.

In 1930, a British Cabinet committee examined two routes from the Kirkuk fields to the sea. The shorter ran to Tripoli, through territory under French mandate. The longer reached Haifa. The estimates were £8.5 million and £11 million respectively. Britain could save on construction and accept a French-controlled outlet, or pay more to reach a port under its own mandate.

The Admiralty’s concern was explicit: “no oil would be obtainable in the event of a war with France”. France might also embargo shipments without explaining why. The Iraqi objection, reported by the British High Commissioner, was to an alignment through Syria.^⁠[⁠4⁠]

The compromise was a fork at Haditha. One branch would run through Syria to Tripoli; the other through Transjordan to Haifa. In February 1931, the Cabinet made its approval conditional on at least half the oil going to Haifa. By 1934, oil was reaching both ports. The arrangement gave the company two outlets and divided its political exposure.^{⁠[⁠4⁠,⁠5⁠]}

War took one of the outlets away. The Haifa refinery stopped in April 1948 amid the fighting in Palestine. The surviving records do not establish a single uncontested date for the last oil through the line. They are clearer about the political refusal. In June, Iraq’s defence minister told the British ambassador that Haifa must be in an Arab state or become a free port. Otherwise the line would have to be relaid through Syria and Lebanon. Iraq could not accept an outlet in a Jewish state.^⁠[⁠6⁠]

Syria now lay across the replacement route as well as the remaining one.

Another company was already seeking passage. Aramco’s Trans-Arabian Pipeline, Tapline, needed Syrian ratification for its shortcut to Sidon. That had not been secured when Colonel Husni al-Za’im seized power in March 1949. His government ratified the agreement by decree in May. The sequence is documented; the claim that Washington engineered the coup for the pipeline remains disputed. The CIA’s own contemporary summaries complained of delays in ratification after the coup.^⁠[⁠7⁠]

For IPC, the choice had narrowed. In April 1950, a Socony-Vacuum director told the State Department that IPC would proceed with a thirty-inch line to Baniyas. Reopening Haifa could not then be discussed, he said. Damascus, he reported, wanted a refinery at the terminus and some $15 million in retroactive transit payments. These were negotiating demands. The first Kirkuk crude was lifted at Baniyas in April 1952, according to the contemporary trade press.^⁠[⁠8⁠]

A State Department memorandum filed in June 1953 recorded a sixty-six-cent difference between the posted crude prices at Ras Tanura and Sidon. The gap had subsequently narrowed to fifty-four cents. Syria and Lebanon treated the difference as a saving created by the pipeline and wanted it shared equally with them. It was their bargaining argument, not an audited measure of profit.

For Kirkuk, the same memorandum recorded the decisive constraint: there were no facilities to export its output through the Persian Gulf instead.^⁠[⁠9⁠]

A pipeline is a foreign-policy commitment buried in the ground. IPC had laid a replacement for Haifa; Iraq had acquired a deeper dependence on Syrian permission. The company could compare the cost of a pipe with the cost of a ship. Baghdad needed an alternative route that actually existed.

· · ·

In November 1956, the assurances given to Moose had failed. The intelligence reports pointed towards the army.

On 4 November, the CIA attributed the sabotage to Egyptian agents. Two days later, it reported a witness to the demolition of the T-2 station who said a Syrian army unit had carried it out. The later account did not withdraw the allegation of an Egyptian role. It brought the Syrian army into the account, while the civilian ministers professed ignorance.^{⁠[⁠10⁠,⁠11⁠]}

The bulletin of 15 November supplied a name. Lieutenant-Colonel Abdul Hamid Sarraj, head of Syrian military intelligence, had threatened Tapline. According to a company official, its line would be blown up unless it agreed to supply Syria’s entire fuel-oil requirements. Sarraj had allowed forty-eight hours for a decision.

The CIA’s accompanying comment recalled his boasts that he controlled the pipelines. It attributed the IPC demolitions to “Syrian army personnel under his orders”. This was an intelligence attribution, not a recovered order. The comment disclosed no source for that particular conclusion.^{⁠[⁠12⁠]}

Current Intelligence Bulletin, 15 November 1956, page 5, item 2, as released by the CIA on 23 October 2019 (C02063769). The blank boxes are the agency’s redactions: the sourcing line that opened the item and a reference after it remain withheld. Public domain.

Sarraj’s reported demand also exposed a difficulty for Syria: the country needed fuel. The CIA subsequently reported a contract with Moscow signed on 15 November. A first delivery was promised at Baniyas in early December. By 21 December, it reported two Soviet deliveries. Even so, the agency warned that Syria faced a fuel crisis early in January.^{⁠[⁠12⁠,⁠13⁠]}

The same December bulletin recorded the condition Syria’s foreign minister attached to a bypass. Work could begin if Washington issued a statement fixing a date for Israel to withdraw from the Gaza Strip. A limited flow of Iraqi oil was technically possible. Its release had become conditional on a withdrawal from territory far from the pipeline. The agency judged that the decision would lie with the Syrian army leaders it held responsible for the demolitions.^{⁠[⁠13⁠]}

Repair would not settle the matter permanently. In January, the prime minister told the American ambassador that the stations could be stopped again, even after repair, if necessary. A few weeks later, IPC paid Damascus the equivalent of $13.72 million in transit dues owed up to November. The payment did not reopen the line.^{⁠[⁠13⁠,⁠14⁠]}

Nor was the explanation confined to Damascus. In February, Nasser told the American ambassador in Cairo that Syria’s refusal reflected an agreement between Egypt, Syria and Saudi Arabia. The surviving account is his description of the arrangement, not its text. His account suggests that Syrian policy alone cannot explain the continued closure.^{⁠[⁠14⁠]}

Baghdad had to plan around a closure it could not end. On 10 February, Nuri al-Said instructed his deputy to seek $238 million in credit from Washington or the World Bank. He estimated that lost oil revenue would reach $28 million by the end of March. It could reach the larger sum if the line stayed shut through March 1958. The CIA still judged Iraq’s finances sound and read the proposed borrowing as protection for its development programme.^{⁠[⁠16⁠]}

Syria authorised repairs on 6 March. Pumping resumed on 11 March. The trade press reported oil reaching Baniyas through the bypass that week, with the destroyed pumps still requiring months of work. Oil could move before the stations were rebuilt, once Syria allowed it.^{⁠[⁠15⁠]}

The effect reached beyond the two neighbours. A March paper for the British Cabinet set out the industry’s forecast for Western Europe’s spring supply. It expected 94 per cent of normal demand to be met under then-current conditions, or 100 per cent with the IPC pipeline open. Those were conditional forecasts, not a measurement of the stations’ share of European oil. The canal was still being cleared. This closure had occurred during the Suez war, and pumping restarted before the canal fully reopened.^{⁠[⁠16⁠,⁠17⁠]}

By August, someone writing to Washington had found a phrase for the mutual exposure. The author, whose name is deleted in the published record, offered personal views later forwarded to Dulles. Keeping the pipelines running, he argued, was evidence that Syria had not wholly moved into the Soviet camp. Closing them would alienate Iraq and Saudi Arabia and improve the West’s position for action. Yet intervention from outside would also put both lines at risk.

“These lines are therefore a kind of hostage in reverse.”^{⁠[⁠18⁠]}

The phrase contained a restraint as well as a threat. Syria could interrupt the oil. Doing so could bring consequences it did not control.

· · ·

Ten years later, Syria used a decree.

The valve as the CIA printed it the morning after the decree: Central Intelligence Bulletin, 9 December 1966, map page, with the Haifa limb marked “Inactive” (CIA-RDP79T00975A009400260001-9, released 27 February 2003). Public domain.

The Ba’ath government wanted higher transit fees and a retroactive increase reaching back a decade. Talks broke down on 23 November 1966. Five days later, an IPC representative argued that the company had reached its limit: a further concession would make shipping Kirkuk oil through Basra cheaper.^{⁠[⁠19⁠]}

Basra’s capacity imposed a limit. The American embassy considered maximum liftings through its existing facilities. Even then, under the assumptions it used, Iraq would receive less than half its previous income. Basra could not yet fully replace the Mediterranean route.^{⁠[⁠19⁠]}

In December, Syria doubled the fee, imposed a surcharge for the back payments it claimed and seized IPC’s assets on its territory. The CIA dated the action to the 8th; the State Department’s published account gave the 9th. From 12 December, Kirkuk oil stopped leaving the Mediterranean terminals.^{⁠[⁠20⁠]}

The company had time. Its pipeline was long amortised, the CIA noted, and its shareholders had other sources of oil. They were under no overwhelming pressure to concede.

In Baghdad, the prime minister was worrying about salaries.

On 23 January 1967, Naji Talib warned the American ambassador that he would soon be unable to pay officials and army officers in full. Iraq’s finances, he said, had been desperate even before the closure.^{⁠[⁠21⁠]}

Official estimates of the daily loss varied over time. The CIA put it at $750,000 a day in December, equivalent to seventy per cent of government revenue. A State Department paper in February used $630,000 a day. Neither figure was a settled account of the final cost.^{⁠[⁠20⁠]}

Syria was also paying to maintain the pressure. Its Homs refinery depended on IPC crude. The February State Department paper estimated forgone transit receipts at more than $2 million a month. In public, the Syrian information minister treated retreat as a political danger, saying a government that yielded would be overthrown by the masses. The closure had made compromise costly for Damascus while making delay costly for Baghdad.^{⁠[⁠21⁠]}

The American embassy in Cairo had doubts about IPC’s case. In December, the embassy in Cairo had asked for a candid appraisal of both sides. It suspected IPC might be reneging on its agreement or reading it too narrowly. That was an embassy’s concern, not a finding that the company had breached its obligations. Cairo was still asking whether Syria had a legitimate claim.^{⁠[⁠26⁠]}

IPC’s parent companies had already agreed, on 23 December, to pay $10.3 million against the Syrian claim for 1966. Other demands were left to arbitration. Exports resumed on 2 March 1967 under a new Syrian convention; a Lebanese agreement followed. IPC’s annual review recorded an average increase of more than fifty per cent in transit and terminal dues across the two conventions. The CIA also recorded a concession to IPC: no time limit on negotiations over back fees.^{⁠[⁠22⁠,⁠27⁠]}

In May, the company advanced Baghdad £13.9 million following the closure. Of that, £5.9 million was recovered from the year’s revenue. Talib’s government fell that same month. The fiscal crisis preceded the fall of the government. The documents cited here do not establish that it caused the fall.^{⁠[⁠22⁠]}

One response had already been suggested. During the closure, the American ambassador suggested offering a moderate Iraqi government the prospect of a new pipeline from K-2 to a Gulf loading terminal. A way around Syria could itself become an instrument of influence.^{⁠[⁠19⁠]}

· · ·

Reopening left Syria with the power to close the line again.

In May 1970, a bulldozer ruptured Tapline on Syrian territory. The CIA’s later account allowed that it might have been an accident. Syria nevertheless withheld permission for repairs while demanding a $50 million advance and higher fees.^{⁠[⁠23⁠]}

This time, the scarcity of ships increased the value of delay. Suez remained closed. A November intelligence estimate compared two routes to Europe. It reckoned that Gulf oil sent round the Cape required five or six times the tanker capacity needed for oil loaded at Mediterranean ports. Tapline’s closure and Libya’s production cuts together had absorbed spare shipping and raised new charter costs. The estimate also judged another politically motivated Syrian interruption of IPC likely within five years. In January 1971, the agency reported agreement to reopen Tapline on higher fees.^{⁠[⁠23⁠]}

Ownership changed the following year. Iraq nationalised IPC on 1 June 1972; Syria announced the takeover of the Syrian section in early June. The CIA considered the actions apparently coordinated. Under the supply conditions then prevailing, it judged that Western Europe would not be seriously affected because alternatives existed. For Iraq, the problem of finding buyers and getting oil to them remained.^{⁠[⁠24⁠]}

Syria’s leverage depended on the alternatives available to Iraq and its customers. Each interruption gave them another reason to seek alternatives.

· · ·

In July 2026, a driver called Abu Saif spent five days getting through the Syrian system to unload at Baniyas. One day was spent on the road, four in the queue. A New Lines reporter found him at the port. A Syrian Petroleum Company official described about five thousand Iraqi tanker trucks circulating through the system, arriving, waiting, unloading and returning.^⁠[⁠1⁠]

The cargo was fuel oil. Crude required a separate correction. On 7 October, S&P Global withdrew a statement that Iraqi crude was being exported through Baniyas. Such exports, it said, “have not commenced”. The 85,000 barrels a day in the earlier account were internal sales recorded in Baghdad on a revenue basis.^⁠[⁠2⁠]

Plans for a pipeline were further ahead on paper. In July, the State Department welcomed Iraqi and Syrian intentions to rehabilitate and reconstruct the old route. The consortium’s release left scope, route, timetable and commercial terms subject to technical studies. In August, two people involved in the project told Reuters that a new line would take about four years and at least $15 billion. These remained estimates and intentions. They were not a construction schedule.^⁠[⁠3⁠]

The queue gives the proposal an immediate appeal. A working line could move oil without that procession of lorries. But the old records describe another kind of waiting: pumps capable of moving some oil, a company ready to proceed, and governments arguing over the permission.

In 1956, a minister who was also an engineer offered to help with the repairs. The obstacle outlasted his offer. What would a new agreement change about who can stop the oil?

Source notes

These notes give the provenance and the limits of the evidence behind each passage. A primary document verifies what its author recorded; an intelligence attribution or a reported conversation does not by itself verify the event beneath it. Where a figure or a date rests on one witness, the note says so. CIA links go to the Internet Archive’s page-for-page mirror of the CIA FOIA Reading Room; the Cabinet papers were opened as scans ordered from the National Archives’ Discovery catalogue and are Crown copyright, quoted under research terms. The argument and the transitions are mine.

1. The driver and the queue. REPORTED CLAIM. Hussam Hammoud, New Lines, 5 August 2026: Abu Saif’s five days (8–13 July) and the figure of about five thousand tankers in the system, from Syrian Petroleum Company official Ali Qabahji. One reporting chain for each; the fleet figure is not daily throughput.

2. The accounting correction. VERIFIED FACT about the correction. S&P Global, 7 October 2026: Baniyas crude exports “have not commenced”; the 85,000 b/d were internal sales recorded in SOMO’s September export data on a revenue basis. It says nothing about fuel-oil movements, which continued.

3. The proposed reconstruction. VERIFIED FACT for the statements: State Department media note, 17 July 2026, checked against two reproductions; UCC Holding consortium release, 18 July 2026, final paragraph. REPORTED CLAIM for the four years and $15 billion: Reuters, 17 August 2026, two unnamed participants, one chain; Chevron said in the same report that refit, expansion and new build were all still options. No financing close or construction schedule exists in the record.

4. The Cabinet routing papers. VERIFIED FACT, read in the scans. CAB 24/211/35, C.P. 136 (30), report of the Cabinet committee on the Baghdad–Haifa railway and pipeline, 1 May 1930: £8.5 million against £11 million, three shillings a ton, £600,000 a year on four million tons, the Naval Staff’s “no oil would be obtainable in the event of a war with France” and the embargo sentence. CAB 23/64/5, conclusions of 21 May 1930: the High Commissioner’s telegram of 13 May, Baghdad “most strongly opposed to the adoption of an alignment through Syria”. CAB 23/66/13, conclusions of 5 February 1931, with CAB 24/219/36, C.P. 36 (31): “bifurcation at Haditha” and the fifty per cent condition, which is the Cabinet’s condition on production, not the wording of the concession. The completion dates are the agreement’s: Hansard, 15 February 1933, vol. 274 cc997–8, on the agreement of 24 March 1931.

5. Construction and first oil. REPORTED CLAIM. Petroleum Times, Review of Middle East Oil, June 1948 (Exeter scan), “The Iraq–Mediterranean Pipelines”: August 1932, 4,878 local and 182 foreign staff, first oil at Tripoli on 14 July 1934 and at Haifa on 14 October. Written with the company’s help; one witness for each figure.

6. Haifa, 1948. REPORTED CLAIM for the dates. FRUS 1949 VI d48, Mattison to Hare, 12 July 1949, reconstructs the 24 April closure and the 4 May Iraqi cut-off a year after the fact; one witness, and the day the refinery stopped is given differently by other official sources, so the month stands alone. A parliamentary answer of 1 June 1948 said to describe intermittent pumping could not be opened and is not used. VERIFIED FACT for the conversation: FRUS 1948 V/2 d392, Douglas to State, 18 June 1948, reporting the defence minister’s words to the British ambassador on 14 June.

7. Tapline and the coup. REPORTED CLAIM for the ratification: FRUS 1949 VI d24, editorial note, Legislative Decree 74 of 16 May 1949, from Damascus telegram 284. VERIFIED FACT for the coup date, two records: Hansard, 6 April 1949, written answer, vol. 463 c184W; FRUS 1949 VI d1132, Acheson to Truman, which does not mention Tapline. CIA Weekly Summary no. 44, 1 April 1949 (”delays in the ratification”), and no. 47, 22 April (”vacillation over the Tapline”). INTERPRETATION: analysts’ summaries show how the agency wrote about the coup, not what its operations side did; the question of American involvement stays open.

8. Baniyas and the back account. REPORTED CLAIM. FRUS 1950 V d18, Funkhouser memorandum of conversation, 25 April 1950, Charles Harding, “Director of Socony Vacuum”: the thirty-inch line, the refinery demand, “retroactive payments amounting to approximately $15,000,000” for fifteen years. Negotiating positions, one witness. First crude lifted at Baniyas, 11 April 1952: Oil & Gas Journal, 5 May 1952, p. 167, one trade-press chain.

9. The posted-price memorandum. VERIFIED FACT about the positions recorded. FRUS 1952–54 IX/1 d294, Eakens, Petroleum Policy Staff, file 887.2553/6–353, which is the source of the 3 June 1953 date; the memorandum prints none. Ras Tanura $1.75, Sidon $2.41 then $2.29, differentials of $0.66 and $0.54; “this saving should be shared equally with transit countries”; “no facilities for the export of Kirkuk output via the Persian Gulf as an alternative”. The memorandum does not say who posted the prices. The gap was the transit governments’ argument, not an audited margin.

10. The cut and the ministers. REPORTED CLAIM. FRUS 1955–57 XIII d336, Moose, Damascus telegram 1040, 3 November 1956: communications cut at 5 p.m. on 2 November; the ministers “claimed not to know where or by whom perpetrated.” The 3 a.m. stop is the CIA’s, Current Intelligence Bulletin, 4 November 1956, p. 11. The denial is reported speech, not evidence of what the ministers knew.

11. Attribution and the bypass. REPORTED CLAIM, each item one reporting chain. CIA Current Intelligence Bulletin, 4 November 1956, p. 11, “Egyptian agents”; 6 November, p. 12, the redacted witness and “a Syrian army unit”; 12 November, p. 3, the damage survey from Beirut reports of 8 November and the superintendent’s 80,000 b/d bypass “if the Syrian government were willing”. The bypass was a limited flow around the wrecked stations, not their repair. Six months to a year: FRUS 1955–57 XVI d554, 303rd NSC meeting, 8 November 1956. Release approvals on the scans are dated 23 and 24 October 2019.

12. Sarraj. REPORTED CLAIM, and an attribution rather than an order. Current Intelligence Bulletin, 15 November 1956, C02063769, p. 5, item 2 and its Comment, released 23 October 2019: the Tapline official is named as the source of the ultimatum; the Comment’s attribution to “Syrian army personnel under his orders” carries no disclosed source. No order, log or named engineer has been found. The same demand reached the embassy on 14 November from the public works minister (FRUS XIII d339, note 2), probably the same company source heard twice. The Moscow contract of 15 November, with first delivery at Baniyas by 3 or 4 December, is the bulletin of 1 December 1956; one witness.

13. The Gaza condition and Syria’s own fuel. REPORTED CLAIM. Current Intelligence Bulletin, 21 December 1956, p. 3: the foreign minister’s condition, the army leaders’ decision, stocks of 32,000 tons against about 600,000 a year, two Soviet deliveries, and the January crisis as a forecast, not a recorded outcome. Asali’s “even if repaired, could be stopped later ‘if necessary’”: bulletin of 9 January 1957, p. 7.

14. Payment and Cairo. REPORTED CLAIM, one chain each. Bulletin of 1 February 1957: $13,720,000 paid on 30 January for transit up to November, and a $700,000 advance on 10 January. Bulletin of 10 February, p. 3: Asali’s “request by Nasr” of 7 February and the Comment that “the decisions will continue to be made in Cairo”. Bulletin of 17 February, p. 3: Nasser’s “agreement between Egypt, Syria and Saudi Arabia”, said to the American ambassador on 14 February. Two leaders’ accounts, not a text of any agreement.

15. Repairs and the restart. VERIFIED FACT for the record, one witness for each date. FRUS XIII d348, editorial note: Damascus 4674 of 6 March 1957 (Asali’s authorisation) and Baghdad 1507 of 12 March (the Iraqi Economic Ministry’s report of pumping on 11 March). Oil & Gas Journal, 18 March 1957, p. 86: the first Iraqi oil in four and a half months at Baniyas through the thirty-inch line, “months to replace the destroyed pumps”, about 220,000 b/d planned through T-1 and auxiliary units.

16. Iraq’s credit and Europe’s spring. REPORTED CLAIM for Nuri’s instruction: bulletin of 15 February 1957, p. 7, which also reads “Iraq’s fiscal position remains sound” and takes the credit as protection for the development programme; no loan or realised loss of those sizes is established. VERIFIED FACT for the forecast, read in the scan: CAB 129/86/16, C. (57) 66, Oil Supplies, 13 March 1957, the two-column table for Western Europe, April–June: 94 per cent “under present conditions”, 100 per cent “with Iraq Petroleum Company pipeline open”. A forecast of supply, not a measure of the line’s share of European oil; the OGJ newsletter of 18 March carries the same 94 per cent from the same industry estimate.

17. The canal chronology. VERIFIED FACT about the statement: Hansard, 25 March 1957, “small vessels can apparently now pass through”, thirty-three-foot ships not before 10 April. A dated forecast, one witness; the point is only that the pumps restarted before the canal reopened.

18. Hostage in reverse. VERIFIED FACT about the argument, not about policy. FRUS XIII d371, letter of 24 August 1957 forwarded to Dulles as personal views, S/S Files, Lot 66 D 123; both the forwarding official and the quoted author are deleted in the published text. FRUS XIII d388, White House memorandum of 7 September 1957, Eisenhower Library, Whitman File: “Oil revenues account for half of the Iraqi income.” The “half” is the memorandum’s figure, not an audited share.

19. The Basra ceiling. REPORTED CLAIM. FRUS 1964–68 XXXIV d196, note 3: talks broken off on 23 November 1966; Baghdad telegram 1014 of 28 November, the company representative’s “cheaper ship Kirkuk oil via Basra”. d197, Strong, Baghdad 1135, 17 December: “less than half as much as heretofore” with Basra liftings “to maximum possible with existing facilities”, and the K-2-to-Gulf pipeline “held out to a moderate GOI as a reward”. The company’s comparison is qualitative; no freight series is supplied.

20. The decree and the losses. REPORTED CLAIM. Central Intelligence Bulletin, 9 December 1966, pp. 7–8, dates the decree and seizure to the previous day and supplies the amortisation, the alternative sources and the Homs dependency; FRUS d196 note 3 gives 9 December; the two dates stand. Export ban from 12 December: IPC’s Review for 1967 and the CIA’s bulletin of 2 March 1967. Bulletin of 19 December 1966, p. 3: $750,000 a day, 70 per cent of government revenue, Arif “had not been told”. The State Department’s 1 February 1967 paper, d202, used $630,000 a day; estimates of one loss, not two confirmations.

21. Payrolls and the walls. REPORTED CLAIM, one witness each. Baghdad telegram 1304 of 23 January 1967, quoted in d201, note 3: “desperate even before pipeline closure”; a warning, not a payroll ledger. d200, Damascus 809, 12 January 1967: Zu’bi “making issue political not economic”, the government “would be overthrown by masses”, posters on 11 January “including walls Embassy and ConGen Aleppo”. d202: Syria’s loss above $2 million a month and the Homs shortage, a State Department estimate.

22. Settlement and aftermath. REPORTED CLAIM for the concession: bulletin of 24 December 1966, p. 6, $10.3 million agreed on 23 December, Shell and BP over their American partners, the rest to arbitration. VERIFIED FACT from the company’s own record: IPC, Review for 1967 (Exeter scan): exports resumed 2 March 1967, the Lebanese agreement “shortly afterwards”, “over 50%” across the two Supplemental Conventions, the £13.9 million advance of May 1967 and £5.9 million repaid. Bulletin of 2 March 1967, p. 6: “knuckled under”. Talib replaced in May: d206, editorial note. INTERPRETATION: the sequence does not establish cause.

23. Tapline, 1970, and the tanker squeeze. REPORTED CLAIM. Central Intelligence Bulletin, 16 July 1970: the bulldozer of 3 May, “possibly by accident”, the $50 million advance demand, the $5 million counter. NIE 20/30-70, 14 November 1970, also FRUS 1969–76 XXXVI d61: five to six times the tanker capacity for Gulf–Cape–Europe against Mediterranean–Europe, a capacity ratio and not a freight ratio; Libyan cutbacks and Tapline named together as the cause of “sharp rises in new charter fees”; the two closures “for political reasons” and the five-year forecast. Reopening on higher fees: bulletin of 30 January 1971.

24. Nationalisation, and the consortium that was never built. REPORTED CLAIM. Central Intelligence Bulletin, 2 June 1972, p. 2: Iraq’s nationalisation of 1 June, Syria’s announcement “today”, the act “apparently” coordinated, Europe not seriously affected “because alternative supplies exist”, a judgment on the supply conditions of that moment. Bulletin of 24 March 1957, p. 7: the companies “tentatively agreed” on a consortium for two lines from the Gulf to the Mediterranean, 1.2 million barrels a day, the first by 1960, “protected by treaty”; a plan, recorded by the agency, that was never executed.

Images. The opening image is the Current Intelligence Bulletin of 15 November 1956, page 5, item 2, as released by the CIA on 23 October 2019 (C02063769); the blank boxes are the agency’s redactions, the sourcing line that opened the item and a reference after it, both still withheld. The map is the one printed in the Central Intelligence Bulletin of 9 December 1966, the morning after the Syrian decree, with the Haifa limb marked “Inactive” (CIA-RDP79T00975A009400260001-9, released 27 February 2003). Both are US government works in the public domain; both are reproduced from the Internet Archive’s mirror of the CIA FOIA Reading Room without alteration beyond cropping.

Hero photograph. American Colony (Jerusalem), Photo Department, July 1933; Library of Congress, Matson Collection, LC-DIG-matpc-15775, item 2019706553. The photograph shows construction on the earlier Haifa route, not Baniyas construction or the 1956 sabotage. Library of Congress catalogue; collection rights statement. The catalogue states: “No known restrictions on publication.” The CIA bulletin described as the “opening image” in the original image note above is reproduced in the 1956 section.

Supplementary source notes

25. The ministers’ meeting. VERIFIED FACT. Moose’s telegram 1040, Damascus, 3 November 1956, 9 p.m., records the meeting at 5 p.m. that day, the ministers’ agitation, their professed ignorance, the earlier protection assurances of 1 and 2 November, and Public Works Minister Jabri’s offer of personal assistance as an engineer. FRUS 1955–57, vol. XIII, document 336, pp. 594–595. Department of State, Central Files, 883.2553/11–356. Single witness for these details; verified as the ambassador’s account, not independent proof of the ministers’ ignorance.

26. Cairo’s reservation about IPC. VERIFIED FACT. Telegram 3385 from Cairo, 16 December 1966, paragraph 6, records the embassy’s suspicion that IPC might be reneging on the existing agreement or construing it narrowly and asks Washington for a candid appraisal. FRUS 1964–68, vol. XXXIV, document 196. NARA, RG 59, Central Files 1964–66, PET 15–2 SYR. Event and document date: 16 December 1966. Single witness; an embassy assessment, not a finding of contractual breach.

27. The concession to IPC. REPORTED CLAIM. The Central Intelligence Bulletin of 2 March 1967, printed p. 6 (PDF p. 7), Notes, Syria item, records IPC’s success in obtaining no time limit on negotiations over back fees for 1955–66. CIA-RDP79T00975A009700120001-1, release approved 18 April 2003. Single witness. The agreement was reported as subject to final Syrian ratification. Its forecast of reopening that day or the next is not evidence of completed reopening; the IPC annual review cited in note 22 supplies the reported restart on 2 March.

Memorandum of telephone conversation, the President and Prime Minister Blair, October 14, 1998, 2:38–2:58 p.m. EDT, Oval Office, CONFIDENTIAL, declassified October 14, 2015. The letterhead and page 5, set together: “I have to keep people jollied up here.”

Part 1 was the way up. Part 2 was the collapse. This is the year after: two refusals, the bill collected, and the second door. Every scene here happened twice. Once in public. Once in the file.

On the morning of Friday, August 14, 1998, the heads of Hong Kong’s three largest stockbrokers were invited to breakfast at the China Club. The invitation had come at short notice. One man was waiting for them: Norman Chan of the Hong Kong Monetary Authority. He asked them to finish their coffee and switch off their mobile phones. Then he walked them across to the Authority’s offices and told them that when the market opened, the government of Hong Kong would start buying it.

That is Chan’s account, written in 2019, when he ran the Authority, and it is memory, like the rest of his morning: half a dozen telephone lines with make-shift recording equipment, and a colleague, Amy Yip, running the room. The part that is not memory is in the minutes of the Legislative Council: the chairman of the securities regulator, the man whose job was to police the market, “was only informed of the Government’s intention in the morning on 14 August.” The exchanges had not been told either. Ten trading days later, on the day the August bets came due, the government had spent HK$118 billion, about US$15 billion, and was, in Chan’s words, “almost being the only buyer in town.”

Two governments broke the rules that year. Hong Kong bought its stock market. Malaysia shut the door on money leaving and fixed its currency. Both were criticized in public. The internal papers are milder: three weeks before the Vice President praised the “brave people of Malaysia” in Kuala Lumpur, the President told the British Prime Minister that “these capital credit controls for the short term can work,” and said why: “because they have the cash reserves.” Seoul and Jakarta, whose reserves could not cover what was falling due, took the program; the two that still had money could choose. The bill for everyone who followed the script went, as before, to people who never signed for a loan, and the countries that had been lectured spent twenty years building a second door.

Ten days in August

The trap Hong Kong was in was arithmetic. The Hong Kong dollar was fixed to the American one, and the price of that promise was that whenever anyone sold Hong Kong dollars, the Authority had to let interest rates rise until it hurt enough to stop them. In October 1997 the overnight rate had touched, in Joseph Yam’s words, “nearly 300%.” A fund that sold the currency pushed rates up. Higher rates pushed the stock market down. A fund that had already sold the market short was paid both ways. A family in a flat in Sha Tin was in the trade too. The rate the banks paid to fund its mortgage sat above 10 percent for months after October 1997, and the banks, by Chan’s account, held their lending rate down for as long as they could. Yam called it the double play, and by August 1998 he had counted four attacks, each costing the city the same way: a spike in the rate every borrower paid and a fall in everything priced in Hong Kong dollars, the profit booked somewhere else. He wrote about it in the South China Morning Post on August 24, ten days in: a handful of hedge funds treating Hong Kong “somewhat like an ATM machine, as others have put it.” He took responsibility: “As a firm believer in such a policy, I am responsible for conducting the intervention.”

The numbers Financial Secretary Donald Tsang gave the legislature on September 8 were the numbers of a siege. Outstanding bets on the index had jumped by more than ten thousand contracts in three days, to over 110,000, “giving strong indications,” he said, “that there was planned and co-ordinated cross-market manipulation.” The government’s answer was to stand on the other side. By Chan’s account the index rose 564 points on the first day.

Then the arithmetic ran in reverse. Yam had put the funds’ bet at 80,000 short contracts, HK$4 billion for every 1,000 points the index fell. Between August 13 and 28 the index rose nearly 20 percent, and every point of it was now a cost. The currency borrowings behind the bet, over HK$30 billion, were costing “around HK$4 million a day” in interest to hold. Two of the four big funds, the regulators later found, had added to their positions while the government was buying. Then the futures exchange raised the cash that big bettors had to post, and open positions fell by a quarter, “over 50% of this reduction” being “concentrated in a small number of accounts.” On August 28, the day the August contracts settled against the index, the government kept buying to hold the index up, and the exchange turned over HK$79 billion. A short contract settled at that level paid the difference in cash; the rest rolled into September, against a government that was still holding. “The manipulators were forced to close out their short positions,” Yam said that November, “in many cases with heavy losses.” That is the government’s account, and the funds’ books are not public. But the numbers are the funds’ own arithmetic, turned around.

I have one more account, and it is mine. That autumn I visited the chief strategist of one of the largest firms on Wall Street, and I asked him straight out what was going on in Hong Kong and what the firms were doing there. He grinned. “Don’t go there,” he said. “It’s been very bloody.” I did not press him. It is twenty-eight years old, from a man I will not name. But it is the only time anyone from that side of the trade told me how August had ended for them, and I took it to mean their side of it, not Hong Kong’s.

Legislative Council, Panel on Financial Affairs, minutes of September 8, 1998: the Financial Secretary’s account of the operation, and the regulator’s evidence that its chairman was told on the morning of August 14.

Two documents by others add to it, one secret at the time. The Federal Reserve’s confidential briefing book for its meeting of September 29, 1998, dated six days earlier, has a page on Hong Kong. The Authority, it says, “spent an estimated $10-15 billion buying stocks in the second half of August, hoping to ensure that speculators lost money,” and “authorities claimed success over speculators.” The staff hedged twice, with “claimed” and with the “so-called ‘double-play’,” and counted the cost in reserves, from “$97 billion at the end of July” to “less than $80 billion.” There is no lecture, and the words “free market” do not appear on the page.

Federal Reserve Board staff, “Recent Developments,” September 23, 1998, prepared for the FOMC meeting of September 29. Released under the Fed’s five-year rule.

The second is the one the funds would have wanted read. In April 2000 the world’s financial regulators, meeting as the Financial Stability Forum, published their inquiry into the hedge funds of 1998. They found that “at end-August, four hedge funds accounted for 50,500 contracts or 49 percent” of the open bets on Hong Kong’s index. And there they stopped, “unable to reach a conclusion,” across the six economies they studied, “on the extent to which manipulation and collusion might have occurred.” Four funds held half the bet against a city of six and a half million people, and the regulators of the world could not say whether they had planned it together. Hong Kong’s government had not waited to find out.

Financial Stability Forum, Report of the Working Group on Highly Leveraged Institutions, April 5, 2000, Annex E: at the end of August four hedge funds held 49 percent of the open interest in Hang Seng futures and options.

In public, Washington was another matter. The State Department’s report on Hong Kong for 1999 put “manipulators” in quotation marks and worried that the intervention “may have affected Hong Kong’s reputation for adherence to free-market mechanisms.” The exit came a year later, and it is the only sale in the series that went to ordinary buyers.

LEDGER · August 1998. Protected: the currency peg, and the depositors and mortgage-holders behind it. Paid: a handful of hedge funds, “in many cases with heavy losses.” Held back: US$15 billion of reserves, moved into thirty-three Hong Kong companies, showing a book profit of HK$35.1 billion at the end of 1998.

The referee

On November 14, 1998, the Legislative Council’s finance panel asked Merton Miller, a Nobel laureate from Chicago, what he thought. A legislator, Emily Lau, put to him the case some were making: “we had made a profit, at least on paper.” The government’s defense by then leaned on New York, where in September the Federal Reserve Bank of New York had called seventeen firms into its boardroom over a failing hedge fund, Long-Term Capital Management, and fourteen had put up the money to save it. “I am more and more sick and tired of hearing about Long Term Capital Management as a justification for what you did,” Miller told them. “It isn’t.” In New York the government had been “a minor traffic cop.” In Hong Kong “the referee was participating in the action,” like the fighter who tells his corner to “keep an eye on the referee because somebody is sure beating the hell out of me.” On the paper profit he gave them a farmyard: “every morning the rooster crows and every morning the sun comes up.” Three officials sat in the room. The minutes record no reply.

Miller’s charge stands whatever the shares did afterward. In Hong Kong the referee bought shares with public money and published the accounts. The New York referee gave the bankers, by the Fed president’s account, “cold coffee and some really terrible sandwiches,” and let them put up the money. Korea in 1997 had been given conditions and the Wall Street fund in 1998 a boardroom; Hong Kong, with reserves of its own, needed neither.

1600 hours

Malaysia had no boardroom to be called into. It had a prime minister who had told the President to his face at Vancouver that the traders “win and we lose,” and a central bank that could not cut interest rates without watching the ringgit fall.

The case for what Malaysia was about to do had already been made, quietly, inside the Fund. The email is timed 9:06 on the morning of Friday, January 30, 1998, and its author, an IMF economist named David Robinson, was replying to a staff argument about a currency board for Indonesia. He had missed the debate, he wrote, and apologized in advance for repeating anyone. Then he did the arithmetic, and the arithmetic he chose was Malaysia’s: three months of 50 percent interest rates “would imply a transfer of about 17 percent of GDP from borrowers to depositors.” He ended with a sentence that sat in the Asia department’s Indonesia file until the file was opened: “if the present situation continues, I would prefer the option of capital controls/ foreign creditors being forced to restructure their debt, which would at least impose some costs on foreigners.”

Email, David Robinson to colleagues in the IMF’s Asia and Pacific Department, January 30, 1998, 9:06 a.m., “currency board option -Reply,” from the department’s Indonesia file, declassified.

A capital control is a border post for money: goods and wages and dividends pass through it, and the hot money that came in on a Tuesday to leave on a Friday is held at the gate. Seven months after Robinson’s email, Kuala Lumpur built the gate, and did it like a bank: by press release, at four in the afternoon, two days running. Tuesday, September 1, 1998, embargoed until 1600, a notice headed “Measures to Regain Monetary Independence”: the offshore market in ringgit closed, foreign money in Malaysian shares and bonds held for twelve months, the measures to “be removed should normalisation in the global financial environment take place.” Wednesday, September 2, 1600 again, a second notice, one sentence long: the ringgit “will be quoted at RM3.80 against the US dollar.” The same afternoon the prime minister sacked his deputy, Anwar Ibrahim, who was jailed that month and appeared in court with a black eye he blamed on his jailers. The black eye is in the court record. Weighing the controls is not a verdict on the government that imposed them.

The Fund’s staff later described how it had been done, in what reads like respect for the craft: “three days” to prepare the rules, and “a very well kept secret.” Outside Malaysia the response was “nearly uniformly negative,” and the index makers took the country off the lists the world’s fund managers buy from.

The gate closed on people as well as funds. On August 31, 197,000 investors, nine in ten of them Singaporeans, held Malaysian shares on Singapore’s over-the-counter market, CLOB, which the new rules shut. The discount on those shares went from 9 percent to 49 in three days. Some 24,000 sold everything they held at that discount and, Singapore’s deputy prime minister told his parliament in October, “would have made losses.” “The government cannot protect investors from such risks.” They were the small savers of the richest city in the region, and they paid first.

I watched the controls from the other side of the table, and they left me uneasy. Every banker I spoke to that autumn was against Mahathir, and so was the Fund. On the money, the street was with him. He had said why to the President a year earlier, in the Vancouver memorandum: “When the IMF requires raising interest rates and presses for more access for imports, it creates added burdens that will have political repercussions.” Jakarta showed where that landed, and the Fund’s evaluators later found that the fiscal tightening in Indonesia and Korea “was not warranted.” The question I still cannot answer is what a lender is trying to do when it arrives at a country’s weakest moment and asks it to cut, in the one season when cutting can bring down order in the streets. The long-run case can be right and the timing can still break a government.

Then the verdicts.

September 11, 1998, the press room at IMF headquarters. Stanley Fischer, the Fund’s first deputy managing director, answered a question on Malaysia like a professor marking a paper. The controls were “the attempt that seems to be being made to cut the country off.” They “may, in the short-run, succeed if you have the apparatus to surprise people and catch them, so to speak, but I don’t believe they are helpful over any sustained period.”

Three and a half weeks later, in Kuala Lumpur, the governor of Bank Negara, Ali Abul Hassan, gave a national congress on economic recovery what reads as the reply, in a speech the bank still keeps on its website. “Doing nothing would mean that Malaysia is leaving its destiny in the hands of others.” Before the controls, he said, policy had been run by officials who “blindly believed in the IMF mantra that a high interest rate regime was good for Malaysia.” The banks, he forecast, would swing from a loss to “a profit of RM1.7 billion,” and his last line was the one a governor is paid to say: “We can see the light at the end of the tunnel.”

At the APEC business summit dinner in Kuala Lumpur on November 16, 1998, the Vice President of the United States rose to give a speech the White House said the President would have given himself. “We hear them today - right here, right now - among the brave people of Malaysia,” Gore said of the “calls for ‘reformasi’,” the rallying cry of Mahathir’s opponents in the streets. On the money, the aim was “not to halt or control the flow of capital.” Malaysia’s trade minister told reporters it was “the most disgusting speech I’ve ever heard.” Mahathir, in the audience, was overheard: “I’ve never heard anything so rude.”

Jollied up

Three weeks before that dinner the President had put it differently to Tony Blair. The two records are on the National Archives’ website, under the government’s classification appeal 2013-090, documents 36 and 38. They are the White House notetakers’ records of the President’s telephone calls with Blair on October 14 and October 25, 1998, each headed CONFIDENTIAL with a line through it, each still carrying its blacked-out boxes, and each stamped with the same declassification date, October 14, 2015, seventeen years to the day after the first call. I have not found the lines below quoted on this subject before.

The first call is about the world’s markets. The President, to Blair: “I think it is very, very serious. Much more than what I say in my public remarks, because I have to keep people jollied up here.”

Memorandum of telephone conversation, the President and Prime Minister Blair, October 14, 1998, 2:38–2:58 p.m. EDT, CONFIDENTIAL, declassified October 14, 2015 (ISCAP 2013-090, document 36).

Eleven days later they are drafting a statement for the G-7, and the President is thinking aloud about a world in which “over a trillion dollars a day” changes hands. Blair: “We need prudent rules.” And then the President says what his Vice President will not say in Kuala Lumpur: “Some don’t feel this way, but these capital credit controls for the short term can work -- just like if the market drops more than 10 percent in a day, we can suspend trading. But they don’t work over the long or medium term. Malaysia and Singapore can do this because they have the cash reserves.” Blair: “I agree.”

Memorandum of telephone conversation, the President and Prime Minister Blair, October 25, 1998, 2:14–2:42 p.m. EST, CONFIDENTIAL, declassified October 14, 2015 (ISCAP 2013-090, document 38). The redaction boxes are the government’s.

The line carries less than it seems. Nothing about the September notices; Singapore, which had no controls, named alongside Malaysia; short term only. And Fischer had conceded the short run himself: controls “may, in the short-run, succeed.” Set side by side, the two are not opposites. The Fund’s man allowed it grudgingly, as a way to “surprise people and catch them”; the President allowed it as a rule of the market, like suspending trading when prices fall 10 percent in a day, and named the condition: “because they have the cash reserves.” Three weeks later Gore, in Kuala Lumpur, did not allow it at all. A year on, the Fund’s staff wrote that the controls “appear to have been effective in limiting outflows.” The private record is the better guide to one thing, the one that mattered: the tool could work for a while, for a country that could pay for it; both men said so, and the speech in Kuala Lumpur left it out. The next fight over money, whoever starts it, will have a file too.

The winter after

The forecast did not survive the winter. On March 31, 1999, presenting the bank’s annual report, Ali Abul Hassan had to read out the year: the economy down 6.7 percent, 83,865 workers laid off, the banks he had forecast would make RM1.7 billion showing “a pre-tax loss of RM2.3 billion.” Reserves, though, had risen by four and a half billion dollars, and the ringgit had not moved since the day it was fixed. That September the Fund’s board wrote its retreat in the grammar institutions use for retreats: the measures “had initially been viewed with considerable skepticism internationally,” and the directors “commended the authorities for using the breathing space well.” A Fund working paper in 2006 found the controls “neither yielded major benefits nor were costly,” and that the stock market had read them “as favoring firms with stronger political connections.” Malaysia went through the crisis anyway. What it avoided was the program. Whether that trade was worth what the connected firms were expected to collect, the file does not say.

LEDGER · September 1998. Protected: Malaysian borrowers, whose lending rates kept falling once the central bank no longer had to defend the currency; the currency itself, at 3.80, for nearly seven years. Paid: foreigners holding Malaysian shares and bonds, locked in, then, from February, taxed for leaving early; 197,000 CLOB investors, mostly Singaporean savers, locked in, and the 24,000 of them who sold at a discount that reached 49 percent. In the file: a President who said the tool “can work.”

Who finally paid

The two refusals are the exception. Everywhere else the program ran its course, and at the end of it there was a sale. A fire sale is a sale where the seller cannot wait. The sellers were governments that had taken over their banks to stop them falling; the buyers were whoever had cash in a year when nobody had cash. Nobody here is accused of a crime. The question is who had paid for what was sold.

Behind every one of those sales was a loan that should not have been made, or taken, on those terms, and the indictment runs both ways. The lenders knew what they were doing: Part 2 showed the Bank of England’s court noting, in September 1997, that the big trading banks faced “remarkably little risk in terms of loss,” and Korea’s creditors ending the year with a state guarantee at a higher rate. The borrowers knew too. Robinson’s sums started from a Malaysia where private credit stood at “170 percent of GDP”; the Fund’s own December table for Korea assumed a billion dollars of bank debt falling due every day. Neither side had built a guardrail: not on how much short money a bank could bring in through a window like Thailand’s offshore banking facility in Part 1, and not, as the Financial Stability Forum noted in Hong Kong, on the size of a bet against a whole market. The shape of it is older than Asia. Credit is made cheap at the center and lent freely at the edge; the profits on the way up are private, and the losses on the way down are collected from the tax base of the country that borrowed. The people in the room lent and borrowed as if the music would not stop, and the people who were never in the room paid when it did.

In Indonesia the state had issued IDR 650 trillion of bonds to hold up its banks, by the case study the Bank for International Settlements’ institute published in 2004, and the agency that took over the wreckage valued what it held at “26% of the face value.” Twenty-six cents on the rupiah. Half of the biggest private bank went to Farallon Capital, “a U.S. investment firm,” in March 2002; the interest on the rescue bonds alone was put at IDR 60 trillion for that year, “a serious burden to the state budget.” Most of the families whose banks had borrowed the money did not repay it; the state budget carried it. In Korea, 168.7 trillion won of public money went into the banks and the deposit insurer. By June 2026, 72.9 percent had come back; what did not is being repaid from the budget, and the last instalment is next year’s. Korea First Bank, scrubbed clean by the state, went to the American fund Newbridge in a deal agreed in 1999, and six years later to Standard Chartered, which announced it had paid 3.4 trillion won, “US$3.3 billion,” “in cash.” In Thailand fifty-six closed finance companies’ loans were auctioned at an average of 28 cents on the baht, half of the core loans going to international investors at 28 cents and a third to the state’s own asset company at 17. The fund that had propped them up was left holding a loss its central bank puts at over 1.4 trillion baht: a 1997 debt now in its third decade, paid down by a charge on every protected bank deposit in the country.

And the people. In 2002 two economists, Peter Fallon and Robert Lucas, put in tables in the World Bank Research Observer what the crisis had done to workers. Unemployment in Korea went from 2.6 percent to 6.8. In Thailand it went from under one percent to more than five. In Indonesia the share of people below the poverty line rose from 11 percent to nearly 14, and to nearly 20 when the prices families actually paid were used. Those are national totals, and they cannot follow any one dollar, but they show who bore the collapse.

I looked for the four people from Part 1’s last ordinary summer whom Part 2 followed into the collapse. Youn Sung Mook, the X-ray technician who slept in the lobby of the hospital that owed him, and Kim Myung Yun, who sold hats from a car for four dollars each and cleared twenty-five cents on the day a reporter stood with him. Zen Zainudin, the singer on the blue metal bench outside the employment office. Warih Wijayanti, the architect who taught school for $40 a month. The record ends where the reporters left them in 1998.

The ledger of the decade, then. Farallon, Newbridge, Standard Chartered and the bidders in Bangkok bought what was for sale, and some of them sold it on. The record says nothing else of them; they were buyers at an auction. The budgets of three countries paid. Indonesia’s carried an interest bill of IDR 60 trillion in a single year; Thailand’s still carries the charge, in its second decade; Korea’s pays the last instalment on the 1997 public funds in 2027.

Conceived in the rooms above

On Christmas Eve 1997, the day Part 2’s banks agreed in New York to roll over Korea’s loans, the Central Intelligence Agency in Washington finished a two-page assessment that would sit under a CONFIDENTIAL stamp for twelve years. “China so far has been relatively unscathed by the East Asian financial crisis,” it begins. “Beijing’s tight controls on capital flows” head its list of reasons the damage was “sharply limited,” and the same controls, with the country’s balance of payments, made its promise not to devalue “more credible than similar statements made by other East Asian governments before they devalued.”

Central Intelligence Agency, “Assessing the Potential Impact of the Asian Financial Turmoil on China,” December 24, 1997, CONFIDENTIAL, approved for release November 30, 2009.

New York’s banks were being talked into staying in a country that had opened its doors; the CIA was telling Washington that the country which had kept them shut was the one that had come through. The next twenty years are the record of what the region did about it.

The idea Washington killed in November 1997 came back as a treaty. In the late summer of that year Japan had proposed an Asian Monetary Fund, a pool of the region’s reserves to lend to its members. Eisuke Sakakibara, who carried the idea, remembered it in 2001 as two months of argument: “Larry Summers was furious,” and “in the end Larry prevailed.” The objection, as he recalled it, was “that it would undermine the International Monetary Fund.” One man’s memory, but the file agrees. A cable listed in Part 2 records Treasury’s man in Tokyo, that November, “concerned with anything outside the IMF,” and eight days later the President’s international economic adviser briefed the press that “the IMF, the International Monetary Fund, ought to be central to any such efforts.” Japan paid anyway. In October 1998 its Ministry of Finance announced US$30 billion for the region, half of it to “take the form of swap arrangements,” and a month later, beside Japan’s prime minister in Tokyo, the President credited “efforts like the Miyazawa Plan.” A year after his Treasury had killed the Asian fund, he was pleased to have the Asian money.

Then the region built what it had been refused, one concession at a time. By 2014 the Chiang Mai swap lines among Asia’s central banks, standing deals to trade currencies, had become a single pool of US$240 billion. The share a member could draw without an IMF program was raised that year, by Bank Negara’s announcement, “from 20% to 30%,” and in 2021 to 40. The BRICS pool signed in Fortaleza in 2014 wrote that year’s split, 30 and 70, into its fifth article. The Fund did not lose the argument of 1997. The people building the rival wrote its name into their treaty.

Except in Beijing. The People’s Bank of China began signing swap lines of its own after the crisis of 2008, in its own currency, and nothing in the published terms refers to the Fund. By the end of 2025 foreign central banks had RMB 94.2 billion outstanding under them, by the bank’s quarterly report. In 2023 Argentina paid the IMF in yuan from Beijing’s line.

Washington reopened its door once the crisis was its own: on October 29, 2008, the Federal Reserve announced swap lines of “up to $30 billion each” for Brazil, Mexico, Korea and Singapore, “fundamentally sound and well managed economies.” Dollars without a program, which Kim Young-sam could not get in nineteen minutes in 1997. The Fed did the same in 2020, and in October 2025 the Treasury signed a swap of up to US$20 billion with Argentina, drawn, by the Congressional Research Service’s account, on the fund a White House aide had been “trying to fudge some on” in January 1998.

A door is where a finance minister goes at eleven at night, reserves running out, and comes back with money. In 1997 there was one, and the people behind it wrote the terms. There are four now. The Fed’s is open for five allies and for others when Washington decides. The Fund’s comes with its program. Beijing’s is in yuan, on Beijing’s terms, and the region’s pool is still 60 percent chained to the Fund.

Which door

And yet. Creditors do sometimes pay now: Ghana’s finance minister told parliament in 2025 of “the 37% haircut on the principal of the Eurobond debt,” one of six such write-downs the Fund’s stocktaking counts since 2020. And on Greece, the Fund’s own evaluation wrote it plainly: delay “provided a window for private creditors to reduce exposures and shift debt into official hands… leaving taxpayers and the official sector on the hook.” The Fund has written down what delay does.

Part 1 asked who holds the paper, when it rolls, and who lends when it stops rolling. Part 2 asked what they want back. The fifth: who has a second door, and what does it cost?

Japan’s line to Bank Negara, signed by the Bank of Japan for the finance ministry, came into force on September 18, 2026, three days after the notice. Up to six billion dollars, from the country whose fund was killed in 1997 to the country that was lectured in 1998. The notice runs to three sentences and a footnote, and gives as its reason “regional financial stability.” Whatever else was said before the signing is in a file that has not opened. Until it does, the rest of us are being jollied up.

In the autumn of 1999, fourteen months after the breakfast at the China Club, the Hong Kong government began selling the shares it had bought in those ten days of August. It packaged them into a fund anyone could buy into, the Tracker Fund of Hong Kong, with a bonus for staying, Chan writes. People lined up to subscribe, an AFP photograph from October 1999 shows. The offer raised HK$33.3 billion, by his account, and “more than 184,000 Hong Kong retail investors” bought in: the public buying, at a discount, what its government had bought to beat the funds. By his figures, the disposals brought HK$140.4 billion plus HK$24.6 billion in dividends on HK$118 billion spent, and the Fund kept HK$51.3 billion of shares. The referee had joined the fight. When it won, the shares went to 184,000 people who had never been in the ring. The bonds Part 1 began with are still rolling.

The file — read it yourself

Hong Kong. Legislative Council, Panel on Financial Affairs, minutes of September 8, 1998 (LC Paper CB(1) 535/98-99) and the special meeting of November 14, 1998 with Professor Miller’s evidence (CB(1)1106/98-99). Joseph Yam, “Intervention true to guiding policy,” South China Morning Post, August 24, 1998, and “Coping with financial turmoil,” Sydney, November 23, 1998, both via the BIS. HKMA, Exchange Fund results for 1998, March 26, 1999. Norman Chan’s recollections, HKMA inSight, September 11 and September 18, 2019; the AFP photograph of the Tracker Fund queue, October 1999, as reproduced by the South China Morning Post, November 12, 2024. Financial Stability Forum, Report of the Working Group on Highly Leveraged Institutions, April 5, 2000, Annex E. IMF, concluding remarks of the 1998 Article IV mission, October 30, 1998. US Department of State, United States–Hong Kong Policy Act Report, 1999. Federal Reserve Board, Greenbook Part 2, September 23, 1998, p. IV-36. William McDonough’s PBS interview, April 17, 2001.

Malaysia. Bank Negara Malaysia, Measures to Regain Monetary Independence, September 1, 1998; Exchange Rate of Ringgit, September 2, 1998; Repatriation of portfolio capital, February 4, 1999; the Governor’s speeches of October 6, 1998 and March 31, 1999. IMF, Stanley Fischer’s press briefing, September 11, 1998; Staff Country Report 99/86, Malaysia: Selected Issues, September 1999; Public Information Notice 99/88, September 8, 1999. Monetary Authority of Singapore, Deputy Prime Minister Lee’s reply to parliamentary questions on CLOB, October 12, 1998. IMF Independent Evaluation Office, The IMF and Recent Capital Account Crises, 2003, executive summary. Vice President Gore’s remarks, Kuala Lumpur, November 16, 1998; the Washington Post, November 17, 1998. The Clinton–Blair telephone memoranda of October 14 and October 25, 1998, ISCAP 2013-090, documents 36 and 38. Johnson, Kochhar, Mitton and Tamirisa, IMF Working Paper 06/51, 2006. The IMF Asia department’s Indonesia file for January 1998 (Robinson email, PDF p. 7) and the Vancouver memorandum of November 24, 1997, both in Part 2’s file.

The bill. Fung, George, Hohl and Ma, Public asset management companies in East Asia: case studies, BIS Financial Stability Institute, 2004. IMF, Staff Country Report 00/21, Thailand: Selected Issues, February 2000, Table 4 and paragraph 73 (a scanned document; printed p. 42). Bank of Thailand, the FIDF. Standard Chartered, RNS of April 15, 2005. Fallon and Lucas, “The Impact of Financial Crises on Labor Markets, Household Incomes, and Poverty,” World Bank Research Observer, 2002, Tables 5 and 7. The Korean, Thai and Indonesian balances today are sourced in Part 2.

The rival. CIA, “Assessing the Potential Impact of the Asian Financial Turmoil on China,” December 24, 1997, CONFIDENTIAL, released November 30, 2009. Eisuke Sakakibara’s PBS interview, May 15, 2001. The Talbott–Haraguchi cable, November 12, 1997 (in Part 2’s file). White House press briefing by Dan Tarullo and Jim Steinberg, November 20, 1997. Bank Negara Malaysia, Enhancement of the Chiang Mai Initiative Multilateralisation, August 7, 2014; Bank of Japan, the amended CMIM, March 31, 2021. Treaty for the Establishment of a BRICS Contingent Reserve Arrangement, Fortaleza, July 15, 2014, as promulgated in Brazil by Decree 8.702 (Portuguese text; the English original was not opened). Japan Ministry of Finance, New Miyazawa Initiative, October 1998. Clinton–Obuchi joint press availability, Tokyo, November 20, 1998. Federal Reserve, press release of October 29, 2008, and March 19, 2020 (the five standing lines are named there). Congressional Research Service, U.S. Financial Support to Argentina (R48780); BCRA, October 20, 2025. People’s Bank of China, Monetary Policy Report, Q4 2025, February 2026 (Chinese). Horn, Parks, Reinhart and Trebesch, NBER Working Paper 31105, on China’s rescue lending. Bank of Japan, September 15, 2026. Argentina’s 2023 yuan payment to the IMF is sourced in Part 2.

Today. IMF, A Stocktaking of the Current International Architecture for Resolving Sovereign Debt Involving Private Sector Creditors, October 2025, Tables 1–2. Ghana, 2025 Budget Speech, March 11, 2025, paragraph 100. IMF, Greece: Ex Post Evaluation of Exceptional Access under the 2010 Stand-By Arrangement, June 2013, paragraph 57.

Standing on shoulders: Paul Blustein’s The Chastening; the Commanding Heights interviews; the Legislative Council’s clerks, who transcribed Professor Miller verbatim.

A note on the records

What this part adds. As far as I can find, the two Clinton–Blair telephone records of October 1998 have not been quoted on this subject before; nor has the CIA’s China assessment of December 24, 1997, been set beside the New York Fed meeting of the same day; nor the Fed staff’s Greenbook page on Hong Kong beside the State Department’s public report. The connections are mine, and readers should weigh them as such. The essay’s claim is narrow: that the private record of the autumn of 1998 was closer to Kuala Lumpur and Hong Kong than the public record was. It does not claim the President endorsed Mahathir, or that the Fund’s staff were wrong about the long run; Fischer’s own short-run concession of September 11 is quoted beside the President’s line, and the distance between them is described as narrow. The President’s phrase was “these capital credit controls,” said of Malaysia and Singapore together, with an explicit limit to the short term; the Fed staff reported that Hong Kong’s “authorities claimed success” and called the double play “so-called”; those hedges are in the text. The Robinson email is a reply in a staff debate about a currency board for Indonesia; its arithmetic is Malaysia’s, and its last sentence is quoted in full. The Wall Street conversation is my own memory of the autumn of 1998; the strategist is not named, the words are as I recall them, and no claim in the essay depends on it. The paragraph on the bankers and the street in Malaysia is likewise my own observation from that autumn; the documented parts of it are Mahathir’s words in the Vancouver memorandum and the IMF evaluators’ finding of 2003.

Norman Chan’s account of August 1998 is a participant’s recollection written twenty-one years later, on the Authority’s own website, and is labeled as such wherever it is used (the 564 points, the HK$79 billion, the HK$118 billion and every Tracker Fund figure are his); the contemporaneous figures come from the Legislative Council minutes, Yam’s 1998 article and speech, and the Exchange Fund’s published results. Sakakibara’s account of the Asian Monetary Fund is likewise a recollection, corroborated on the US position by the Haraguchi cable and the Tarullo briefing. The Clinton–Blair memoranda were prepared by White House notetakers; the redactions are the government’s; both records carry the Interagency Security Classification Appeals Panel’s stamp, with the declassification date of October 14, 2015, on their first pages. The Fund’s Malaysia report is a staff paper; the board’s notice is the institution’s view. The BRICS treaty is read from the Portuguese text promulgated in Brazil; the treaty says the English text is authentic, and I have not opened it. The Fallon–Lucas figures use different national definitions of poverty and unemployment and are reported as their tables print them; the Indonesian 19.9 percent is their re-deflation with household-survey prices. The Thai auction, Indonesian bond and Korean sale figures are the amounts the cited documents state; none is a measure of any buyer’s profit, and no buyer named here is described as doing anything other than purchasing what a government sold. The “second door” figures are reported, dated amounts from the linked releases; they describe funding arrangements and do not assess any country’s condition as of publication. Nothing here is a forecast or a recommendation.

Notes and sources

1. The double play: the mechanics are Yam’s own description (August 24 and November 23, 1998); “four attacks” is his count, the “third and fourth attacks in June and August” following two earlier ones. The “300%” overnight rate is his figure for October 23, 1997 (”almost 300%” in the Financial Secretary’s account to the panel). HK$118 billion, 564 points and HK$79 billion are Chan’s 2019 figures; “about US$15 billion” is Yam’s, the State Department’s, and within the Fed staff’s “$10-15 billion.” The book profit of HK$35.1 billion is from the Exchange Fund results for 1998 (March 26, 1999). Tsang’s “over 110,000 contracts” is open interest, all contracts outstanding, not short positions alone. The funds’ costs are Yam’s Sydney figures (80,000 short contracts; HK$4 billion per 1,000 points; “in excess of HK$30 billion in currency borrowings, at an interest cost of around HK$4 million a day”); “nearly 20 percent” is the Fed staff’s figure for August 13 to 28; the margin surcharge and the fall in open interest from 107,000 to 82,000 contracts are the futures exchange’s evidence to the panel (para. 30); the two funds that added positions during the intervention are the FSF’s finding (Annex E). “Heavy losses” is Yam’s claim; no fund’s accounts for the episode are public. Hang Seng Index futures settle in cash against the index, not by delivery of shares; the August 28 turnover (HK$79 billion) and the “more than 100,000” contracts rolled into September are Chan’s figures. The Sha Tin family is an illustration; the mechanism is Chan’s: one-month HIBOR, “the key benchmark for the funding costs of banks for their mortgages,” stayed “over 10%” for months after October 1997, and the banks “held back the increase in prime rates” for as long as they could.

2. The regulator: the “morning on 14 August” line was given by the SFC’s acting chairman, Laura Cha, to the panel on September 8, 1998 (para. 25); the same minutes record that the operation fell “outside the jurisdiction of SFC” under section 66 of the Interpretation and General Clauses Ordinance (para. 26), and that the exchanges and the clearing house “were not informed of Government’s plan in advance” (para. 18).

3. The FSF’s numbers: “49 percent” and “50,500 contracts” are end-August figures for futures and options together (Annex E, printed pp. 130–131). The group’s “unable to reach a conclusion” is its finding across the six economies it studied (§46), not Hong Kong alone. The Greenbook’s silence on the words “free market” was checked against the page’s text.

4. Miller: the “traffic cop” and “referee” analogy he attributes to a colleague, Professor Chen; Emily Lau’s “at least on paper” was put as the view of “some people,” before the LTCM exchange. Three officials attended; the minutes record no government response to Miller. McDonough’s account gives seventeen firms called in and fourteen that “did it.”

5. Malaysia’s numbers: the RM1.7 billion forecast (”expected to make”) and the RM2.3 billion loss are the Governor’s own figures, five months apart; the March 1999 speech adds that excluding three institutions the system made a pre-tax profit of RM793 million. Reserves (US$21.7 billion to US$26.2 billion), GDP (−6.7 percent) and layoffs (83,865) are from the same speech; the fall in lending rates began before the controls (the maximum base lending rate was already down to 8.9 percent at end-August 1998) and continued after them. The exit levy of February 4, 1999 replaced the twelve-month rule. The Fund’s “effective in limiting outflows” is SCR 99/86, para. 29; “three days,” “very well kept secret” and “nearly uniformly negative” are in its section C. The peg at 3.80 was lifted in July 2005 (background). Anwar Ibrahim’s black eye is from the Washington Post of November 17, 1998. CLOB: the figures (197,000 investors at August 31, 1998, 90 percent Singaporean; 24,000 who sold all their Malaysian shares; the discount widening from 9 percent on August 28 to 49 percent on September 3, averaging 42 percent when trading reopened from September 9 to 15) and the two quoted lines are from Deputy Prime Minister Lee Hsien Loong’s reply to Parliament, October 12, 1998, published by the Monetary Authority of Singapore. “Was not warranted”: the IMF Independent Evaluation Office, 2003, executive summary: “the initial tightening of fiscal policy in Indonesia and Korea was not warranted.” The indictment paragraph: “remarkably little risk in terms of loss” is the Bank of England’s Court minute of September 16, 1997 (Part 2); the state guarantee at a higher rate is Korea’s January 1998 creditor deal (Part 2, note 1); “170 percent of GDP” is Robinson’s email; the billion dollars a day is the assumption in the IMF staff’s Korean reserves table of December 8, 1997 (Part 2, Exhibit E), a projection, not a record; Thailand’s offshore banking facility is in Part 1; the absence of position limits is the FSF’s bracket. The sentence on credit made cheap at the center and losses collected from the tax base is the author’s reading, not a finding of any cited document.

6. Gore’s speech: the text quoted is the prepared text as hosted by Mount Holyoke College; the Washington Post reports the delivered line as “calls for democracy in many languages -- people’s power, doi moi, reformasi. We hear them today -- right here, right now -- among the brave people of Malaysia,” describes reformasi as “the rallying cry for opponents of Prime Minister Mahathir Mohamad,” and attributes Mahathir’s remark to Reuters.

7. The calls: Blair’s “We need prudent rules” and the President’s “over a trillion dollars a day” are on page 6 of the October 25 record. Korea’s sale: Standard Chartered’s price (”Korean Won (KRW) 3.4 trillion (US$3.3 billion), in cash”) is the buyer’s own announcement, which names no seller; Newbridge’s 1999 agreement is background. The Lone Star arbitration over Korea Exchange Bank, and its reported annulment in November 2025, are left out of this part; the file for them is in the research ledger.

8. Indonesia’s figures (IDR 650 trillion; 26 percent in IBRA’s 2000 report; Bank Central Asia, the largest private bank, to Farallon; the IDR 60 trillion estimate) are from the BIS/FSI case study, pp. 8–14, as read from the Yale School of Management’s hosted copy. Thailand’s: IMF SCR 00/21, Table 4 and para. 73, read from the page image for this essay; the 1.4 trillion baht headline is the Bank of Thailand’s own, whose page also gives the 0.46 percent charge on protected deposits; the balance outstanding is sourced in Part 2. Korea’s: 168.7 trillion won and the 72.9 percent recovered by June 2026 are the Financial Services Commission’s figures (July 21, 2026); the last instalment is the FY2027 budget proposal of September 1, 2026, which in the government’s words completes the repayment; both are in Part 2’s file. Unemployment: Part 2 gave Korea’s rise as 2.6 to 7.0 percent (annual averages, from Korea’s SEC filing); Fallon and Lucas’s table gives 6.8.

9. The rival: the Chiang Mai Initiative and its 2010 pooling are background, widely recorded; the load-bearing figures (US$240 billion; 20 to 30 percent) are Bank Negara’s 2014 release; the 40 percent is the Bank of Japan’s 2021 notice. The BRICS split is Article 5(c) and 5(d) of the treaty, read from the Portuguese text promulgated in Brazil. The People’s Bank of China’s swap network dates from 2008–09 in the accounts cited in Part 2; RMB 94.2 billion is its own figure for balances drawn by foreign monetary authorities at end-2025 (Q4 2025 report, section 10). The Miyazawa page on the Ministry of Finance’s site is undated; October 1998 is the initiative’s announced month. The Fed’s 2020 lines are its release of March 19, 2020; the Treasury–Argentina swap is the BCRA’s release of October 20, 2025, its funding source the Congressional Research Service’s. Argentina’s yuan payment to the IMF in 2023 is sourced in Part 2 from press accounts and Yale SOM; it was a bridge, repaid from the IMF’s August 2023 disbursement.

10. Today: the IMF stocktaking’s count is its own (Table 2, six external bond restructurings 2020–2024); the Greece evaluation is quoted from paragraph 57. The Tracker Fund figures (HK$33.3 billion raised; 184,000 investors; HK$140.4 billion proceeds; HK$24.6 billion in dividends; HK$51.3 billion retained; the loyalty bonus, one free unit per twenty held for a year and one per fifteen for two) are Chan’s 2019 figures and are attributed to him in the text; the Exchange Fund’s published profit is the only independent number used for the disposal. The Bank of Japan’s release of September 15, 2026 gives the effective date as September 18.

White House memorandum of conversation, November 27, 1997, 2356–0015 EST, Camp David. CONFIDENTIAL, then DECLASSIFIED; the round stamp reads “Dep Sec Has Seen.”

Part 1 was the way up. This is the collapse of 1997–98, Thailand to Korea to Indonesia, told from documents that were secret when they were written and from the streets they were written about. Korea makes the last payment on it next year. Two ledgers run through it: what was protected at the top, and what was paid at the bottom. Every scene here happened twice. Once in public. Once in the file.

At 11:11 on the night of December 8, 1997, an economist at the International Monetary Fund faxed a two-page memo to the head of its Asia department. It was about the money to keep Korea out of default. Borrowing it, the memo said, “would involve a substitution of sovereign debt for private debt.” In plain words: the public would take over what the banks were owed. And that was nothing new, the memo went on, “since such a substitution was already envisaged in the original baseline projection.”

The money had not yet moved. The bill had already been assigned.

Here is what this part of the story shows. The rescue of 1997 was built to make the lenders safe. The banks that had lent the dollars got new loans, a government guarantee and a higher rate. The people who had never borrowed a dollar paid for the collapse: with their jobs, their savings, their wedding rings, and thirty years of taxes. The substitution was not an accident of the crisis. It was in the plan, and the people who wrote the plan wrote it down.

This is the story of how that sentence came to be written. And of the people who were never in the room.

Four minutes to midnight

At 11:56 p.m. on Thanksgiving night, November 27, 1997, the President of the United States telephoned the President of South Korea from Camp David. He called to tell him his country had about a week.

I read the transcript on the State Department’s own website. It was released after a lawsuit. It is stamped CONFIDENTIAL, then DECLASSIFIED, and then, in a round date-stamp from December 1997, Dep Sec Has Seen. As far as I can find, no book or newspaper has quoted it.

White House memorandum of conversation, Clinton–Kim Young-sam, November 27, 1997, 2356–0015 EST. Released by the State Department in FOIA litigation, October 2019. The circular stamp reads “Dep Sec Has Seen.”

“We have been informed by your financial team and by the IMF that your financial situation is extremely grave,” Clinton begins, “and that you face the prospect of default in a short period of time, perhaps by the end of next week.” Kim’s government had asked Washington and Tokyo for a bridge loan: cash to last three weeks. The answer, from the most powerful man on earth to an ally, is no. It “will not work… the money will all be exhausted in a matter of a few days so we could all lose some money.” The call lasts nineteen minutes. At one point the President asks Kim to speak up. He can barely hear him. Then comes the whole year in one sentence: “We can put together the necessary backing if it is clearly a part of an IMF package.”

The next day he called Tokyo. The cable of that call went out marked NODIS and NOFORN: no distribution, no foreign eyes. Even $20 billion, Clinton said, “would be swamped in three weeks.” Hashimoto drew the real map of power: “President Kim has not given much jurisdiction to his finance minister so it will be between you and me, Mr. Mitsuzuka, Mr. Rubin and Mr. Summers to advance the case.” Five men.

Months later, a different payment failure turns up in the reporting from Seoul. An X-ray technician named Youn Sung Mook, 32, unpaid since February, is sleeping in the lobby of the hospital that closed on him.

The distance between those two records is the subject of this story. Whose promises did the rescue make safer? Whose income was left exposed?

The reserve that could not be spent

It starts, in July, with a number that isn’t there. On the morning of July 2, Thailand lets its currency go. The baht falls from 24.25 to the dollar to 27 and 28 by early trading. That is the number the papers print. The number they can’t print is the overnight interest rate, because there isn’t one. Banks will not lend to each other until someone knows what money is worth. The governor of the central bank later told an inquiry about the spring meeting where the reserves were first thrown into the defense. The New York Times reported his words: “Everyone panicked, and some even cried.”

The big funds did bet against the baht, and they made money. But the academic record shows their combined bet fell from about $5 billion to under $2 billion within July. The people who stayed for the whole year were the ones who could not leave.

The sealed record adds the first twist. Two days before the IMF board approved Thailand’s rescue, Australia’s director filed a statement that “Thailand’s net reserves are about half the published gross reserves.” He proposed that disclosure be delayed until November. (It was not. The forward book, the reserves already promised away in currency contracts, was published within weeks.)

LEDGER · Summer 1997. Protected: trading desks and macro funds, with their profits on the baht, out by August. And the major trading banks with business in South East Asia: by the Bank of England’s own Court minutes of September 16, “remarkably little risk in terms of loss, and quite a few had made money out of volatility.” Paid: fifty-eight suspended finance companies, their depositors, and a vice-president named Kittisak, who traded the BMW for a Mazda and a $120-a-month room.

In Seoul, Youn Sung Mook’s hospital will stop paying him in February. This month, that is seven paychecks away.

Ten days

The mechanism takes one paragraph. A country that has borrowed short-term dollars must borrow them again every few weeks. Bankers call it “rolling over” the loans. When one Asian borrower fails, every lender looks at every similar borrower and quietly declines to renew. Refusing to renew a loan is not an attack. When everyone does it at once, it is. A company can own factories and still lack the dollars it owes on Monday. Its bank can be solvent and still hoard cash for its own creditors.

On Monday, November 17, the Bank of Korea stops defending the won. By Wednesday the won is hitting its daily limit ten minutes after the open. The finance minister has been replaced. His successor tells the press: “There is no need for financial help from the IMF.”

The same Wednesday, the CIA’s National Intelligence Daily, Top Secret, released in 2009, told the President’s people something else. Seoul must roll over “about $6 billion monthly.” The rescue packages Seoul and Tokyo were hinting at “may be insufficient.” The brief measured the danger in the currency it understood: US growth “cut about 1 percentage point,” the US current-account deficit “$45 billion” wider, world trade “reduced by nearly $400 billion a year.” The only workers in its three pages are American: “rising US wages and a taut labor market.”

The CIA’s National Intelligence Daily, Wednesday, November 19, 1997: Top Secret struck through, released February 2009.

On the night of the 21st the new minister reads the request out on live television. Six days later comes Thanksgiving, and Camp David.

How is a nation’s economic policy rewritten? In a hotel, quickly. A three-man IMF advance team lands on a Sunday. Korean television counts fourteen Fund staff in all. The mission works around the clock for seventy-two hours. On December 3 Camdessus lands at 7:35 in the morning. The signature comes at 7:25 that evening. Ten days, from the mission’s arrival to a program that commits the world’s eleventh-largest economy to rewrite its banking, labor, corporate and capital-account laws.

That evening the two presidents speak again. The cable of the call has Kim tell Clinton what his country is saying: “the Korean news media is describing today’s negotiations with the IMF as a very shameful one. They think it is very humiliating, almost comparable to Japanese colonial rule. However, I think that there is only one choice.” Again the transcript records the American president asking him to speak up. Clinton asks him to get the three presidential candidates to commit in writing. Kim: “They were rather reluctant.” He would get their signatures anyway.

Dep Sec Has Seen

The board approved Korea on December 4 “in restricted session.” That is a closed meeting, with the debate kept out of even the not-for-public-use minutes. What it was about, I can tell you from a different file: the Asia department’s own crisis file, released in 2019. It holds the December memos, stamped STRICTLY CONFIDENTIAL or SECRET and then DECLASSIFIED. In place of four of them, it holds withdrawal notices: “Korean Loan” (Fischer to the Managing Director, December 2), “Meeting with Executive Directors,” “Indonesia and Korea” and “Korea – Contingency Planning” (all December 9, the last one SECRET). Those four stayed back. What follows is what the Fund let out.

Korea’s usable reserves in December 1997, as the IMF’s staff projected them on December 8, with this chapter’s date-stamps marked. Drawn from the staff table below: the 30 percent rollover scenario, with the Fund’s two instalments as the only money coming in.

December 8 — $9.2 billion, the Fund’s first instalment in. The date-stamps in this chapter are the Fund’s own figures for Korea’s usable reserves, the dollars actually on hand, as its staff projected them that day. A staff economist wrote a memo he headed “Korea: What Can Markets Calculate?” and did what he called “the work of a moment.” Usable reserves had been $6 billion on December 2. If the banks rolled over only 30 percent of what fell due, the table he attached ran day by day to New Year’s Eve: $0.4 billion. The Fund’s two instalments were pencilled in as the only money coming in. There was one day off, December 25, when he assumed no loans would fall due. I found the table on page 140 of the file. It is the most frightening page in this series.

The countdown as the Fund’s staff projected it on December 8: $7.5 billion at end-November, $0.4 billion by New Year’s Eve if only 30 percent of loans were rolled over. A projection under stated assumptions, not a record of what happened. From the Asia department’s crisis file, marked STRICTLY CONFIDENTIAL, declassified 2019.

December 9 — $8.5 billion. In open session, the Dutch director asks whether Korea’s short-term debt is the staff’s $65 billion or the press’s $100 billion. The head of the Asia department answers “probably $65 billion… plus about $50 billion” owed by Korean banks’ overseas branches. The minutes read like a heist discovered in progress. Downstairs, a memo for a two o’clock meeting: reserves “low and hemorrhaging… virtually exhausted by the end of the year.” A colleague had already done the arithmetic. It is the fax this story opened with, sent at 11:11 the night before, with no soft words: borrowing to avert default “would involve a substitution of sovereign debt for private debt. However, this does not involve anything new since such a substitution was already envisaged in the original baseline projection.” The public’s debt for the lenders’ debt was not an afterthought of the rescue. The staff’s own projection had assumed it from the start. In plain words: Korea would borrow from governments to repay the banks. Korea’s taxpayers would then owe the governments. To be fair to the man who wrote it, he was arguing for Korea. The alternative, he wrote, was more adjustment through even higher interest rates, and that “is likely to be costly.” He was choosing the lesser pain. He was also recording who would carry it.

The fax at 11:11 p.m.: “a substitution of sovereign debt for private debt… already envisaged in the original baseline projection” — and, two bullets down, the argument for it: further adjustment “is likely to be costly.” Same file.

December 12 — $6.4 billion. Kim telephones Clinton with the trophy: “I have signatures of all three candidates.” The transcript preserves the American answer to Korea’s plea for more money now. “Some of our people in the Treasury Department and in large investment houses here in America who have a big stake in what is happening in Korea,” Clinton explains, “are afraid that if we provide more money now it will raise doubts in the international financial markets of the adequacy of the current program.” Read it twice. The fear that counted was the fear of the people with a stake.

Five days later the Fund’s first review of Korea, STRICTLY CONFIDENTIAL, carried the sentence that matters more than anything anyone joked about that winter: “Domestic financial markets were virtually paralyzed last week.”Compliance had not brought the dollars. A company whose bank stops lending can have orders to fill and no way to make Friday’s payroll.

In Seoul, the hospital where Youn Sung Mook takes X-rays prepares its March payroll. He will not be paid.

[Laughter]

December 16 — $5 billion. The Federal Reserve met. Its transcript was sealed for five years. The committee was told that “the exchange markets for the won and the baht have disappeared.” The chairman reached for an image, “the tide coming in for the first time in 20 years, and we find all the junk on the shore,” and then a joke: the crisis could turn on a dime, but “it cannot turn on a baht.” The stenographer wrote down what came next: [Laughter]. In public, nothing.

That night, at 9:22 in the Residence, Clinton took a call from Tokyo. Hashimoto had just met the leaders of ASEAN, China and Korea. The transcript has him report “very strong reservations on the part of ASEAN leaders” about the IMF, and “some sentiment for meetings without the Americans present.” Clinton’s reply is the rule of the year in one line: “there is not enough money in U.S. and Japan combined to stop a run on their currencies unless they put their houses in order.”

December 18 — $3.6 billion. Korea elected Kim Dae-jung, the dissident it had once jailed, president of a country with days of dollars left. In his home region of Cholla thousands danced and wept in the streets. That night Clinton called him. The transcript shows the American president coaching the winner on his first words: reaffirm the IMF deal “before the markets close today.” At the Fund, the same week’s minutes record reserves “coming down toward zero.” They also record the American director’s private verdict on Korea’s guarantee to its foreign banks: “the central bank was paying out at the back door to cover all the rollovers.”

December 24 — $3.2 billion. On Christmas Eve the banks were asked. William McDonough, president of the Federal Reserve Bank of New York, called “the major financial institutions of the world” to a room at 33 Liberty Street. It happened “in this very room,” he said there in 2001. He put it to them that no more public money would go into Korea “if the private sector was pulling its money out.” His method, as he described it, was to tell bankers that “there is no public-sector money to solve the problem; the taxpayer is not going to do this.” The same day the Treasury Secretary, Robert Rubin, formerly co-chairman of Goldman Sachs, announced that the G-7 would advance its money “in the context of a significant voluntary extension of the maturities of existing claims by international bank creditors.” The banks agreed to roll over. The won jumped 23 percent in a session. The taxpayer McDonough meant was American. In March, 134 foreign banks swapped $21.8 billion of short-term loans for new one- to three-year paper. The new paper carried Korean government guarantees and margins of 2.25 to 2.75 points over LIBOR, the rate banks charge each other. A promise from Korea’s state, for a fee, to the institutions that had declined to renew. The banks got a promise. The workers got a payment date. Five days after Christmas, the sealed minutes scheduled Korea’s money like an allowance: sixteen dated instalments, out to November 2000. At bank counters across Korea a gold collection began. Wedding rings, baby rings. Seventeen tons in the first week, and by the widely reported final count some 3.5 million people and about 227 tons. The gold went in one door. The guaranteed paper went out the other.

LEDGER · December 1997. Protected: Morgan Stanley Dean Witter, whose foreign-exchange trading revenues “increased 196% in fiscal 1997” on volatility “including Thailand’s baht.” Chase, with $5.4 billion of Korea exposure and “record corporate finance and loan syndication fees” of $1.14 billion. Korea’s creditors, guaranteed by the state for a fee. Paid: 3.5 million people in gold. Unemployment, 2.6 percent to 7.0. Suicides up 42 percent in a single year, the male rate by 45. Koreans named the era after the acronym: IMF price, IMF food,and the one that hurts, IMF papa, a father who has lost his job.

Who would write the rules

The rules had been asked for. Three days before Camp David, at a breakfast in Vancouver, Mahathir told Clinton: “The only ones who have benefited are the traders who have earned $5 billion while we have lost $30 billion. They win and we lose. We need to regulate currency traders.” The memorandum, SECRET, has the President answer that in the United States “currency trading has become a profession.” And then: “I promise I will follow up on this.” Goh Chok Tong confided that Singapore was “quietly buying Indonesian rupiah,” and had to do it quietly “because the public would question the government’s action.” What Washington followed up with is in the cables.

White House memorandum of conversation, Vancouver, November 24, 1997, SECRET, released in full: “They win and we lose.”

Washington also considered what the crisis made possible. On December 12, the day Kim reported the three signatures, a cable went from the State Department to the Seoul embassy “in a specific request from Treasury Deputy Secretary Summers.” Washington wanted Korea, at the WTO’s financial-services deadline in Geneva, to bind the IMF’s conditions into a permanent trade treaty. Its audience: “over fifty of the major international financial institutions — Korea’s key audience — are now in Geneva observing the WTO negotiations.” And it records what happened when the Americans handed Seoul a marked-up version of Korea’s own offer: “Korea angrily returned a copy of our suggestions, noting that this way of providing U.S. ideas offended its national dignity.” A White House email that New Year’s Eve flagged the argument: the crisis “is in fact opportunity to pursue trade liberalization. IMF conditions best leverage we have.”

Canberra read the same package the other way. A minute of Australia’s National Security Committee, March 17, 1998, SECRET AUSTEO, records “Australia’s actions aimed at modifying aspects of the International Monetary Fund (IMF) package likely to threaten social stability.”

Cabinet Minute, National Security Committee, March 17, 1998, SECRET AUSTEO. National Archives of Australia, A14370, JH1998/85.

London said the quiet part in its own boardroom. On January 21, 1998, the Bank of England’s Court, its board, heard Mervyn King explain why the lenders had lent so cheaply to Korea: “in effect they were lending to the IMF or the G10.” A director, Sir Colin Southgate, answered that there was “a case for refusing to bail them out if they would not open their doors. He approved of the IMF using its leverage to force liberalisation.”

And the system billed for its work. A World Bank folder from February 1998 is titled *Options for FY98 Net Income*. Inside it, the Bank’s acting treasurer tells Wolfensohn that “the U.S. and UK representatives are keen to ensure non-standard pricing” for Korea’s next loan, and recommends he “stand firm.” The approved terms carried a 1.5 percent charge “financed out of loan proceeds.” That is $30 million deducted before Korea saw the money, in a document that lists, among the reforms supported, “the recent legislation legalizing worker layoffs.”

The rescuers were not one mind. They were one system. Its fees, spreads and market access were settled in the same weeks the country receiving the money was drawing it, instalment by instalment.

Their defense is in the file too. Clinton, to the Asian leaders in Vancouver, citing Mexico: “the cost of doing it was far less than the cost of not doing it.” He may have been right. A default in December 1997 would have reached the same hospital and the same sales team first, and faster. The Fund’s economist made the same case in the fax of December 8: financing the shortfall was the lesser pain. The file cannot settle whether a gentler rescue was possible, or whether the banks could have been made to share sooner. It can show who was asked to carry which risk. The banks carried a guaranteed loan at a higher rate. The public carried everything else.

The photograph

The file had seen Indonesia coming. On November 7, a week after the first IMF program, seven American executives had lunch with Talbott in Jakarta. Chase Manhattan’s country officer was among them. They told him that “considerable social unrest is almost certain” around the March presidential selection. They also urged Washington to restore military training for Indonesia’s armed forces. In Vancouver on November 24, in a memorandum marked SECRET, Clinton told Suharto he wanted to support him “with military equipment and training, with the understanding that it will not be used to put down unrest.” And, “as a friend,” that “you will have an incredible legacy when you complete your work.” Suharto asked for an emergency fund, because “if we rely on the IMF alone, we prolong the suffering.” The work was completed for him six months later.

Clinton to Suharto, Vancouver, November 24, 1997, SECRET: “I think you will have an incredible legacy when you complete your work.”

Indonesia’s turn came in January, and it came twice. In public: the rupiah through 10,000. Jakarta’s shelves empty by noon. A Foreign Ministry spokesman insisting the people “keep on moving around, laughing, smiling.” A general declaring the situation “safe and stable” on the day a photographer framed hundreds queueing for cooking oil. In the file, that same January 8, Indonesia’s own representative at the board put a two-page warning into the record, stamped NOT FOR PUBLIC USE. Real interest rates were “punitive.” It was “by no means clear that it is worth throwing the economy into a deep recession in order to prevent further depreciation of the rupiah.” The “welfare of poor people,” if not addressed, “could lead to social and even political unrest.” The board took Indonesia in restricted session. Four months later the unrest arrived.

On January 15 came the signing, and the photograph Part 1 promised: Suharto bent over the paper, Camdessus above him, arms crossed. The file has the second version. At 11:27 that morning, Washington time, Suharto telephoned Clinton: “I believe it appropriate to ask you to make a statement to the world.” Clinton answered, “I will release a statement saying you called me.” Three days earlier, by phone, Hashimoto had told Clinton he had spent “more than 14 minutes” telling Suharto “that he should only say that Indonesia would abide by the IMF agreements and not mention extra things.” He had also described Japan’s own bank rescue that week: thirty trillion yen of public funds; “depositors will be protected in full.” Clinton had called it “a very constructive step.” The banks got a promise. Japan’s depositors got a promise. Indonesia’s poor got a warning, in restricted session. A month later, in the declassified transcript of his February 13 call, Suharto handed Clinton the sentence this series has been building toward: “The Indonesian people see the IMF as a savior too late.”

In Jakarta, a karaoke club called New Mugen closes. A man who sang there for six years starts visiting the employment office. In Seoul, Youn Sung Mook works March without pay, and April.

Seven days

Fuel prices rose at midnight on May 4. Two weeks of unrest followed. A noodle vendor with six children, Yayah Syamsiah, gave the year its line: “Who will listen to us? Ordinary people are crying.” On May 12, at Trisakti University, as the rain cleared toward dusk, security forces shot four students dead. A boy in a white headband told a reporter, “They shot my friend in the back of the neck.” The next evening a US defense attaché cabled what he had heard from a high-ranking Indonesian officer. The report is marked CONFIDENTIAL and “not finally evaluated intelligence.” The troops “were not issued live ammunition but somehow between leaving the barracks and arriving at the demonstration some of them obviously acquired some.” It ended: “It is possible that there are factions within ABRI who are trying to promote chaos.” Jakarta burned. By the 20th, soldiers on a tank beside the stock exchange were waving busloads of students on toward the occupied parliament with a thumbs-up. That Monday night, by the Post’s account, a minister who had to carry the message asked the others, “Please pray to give me the strength to say this to the president.” A scholar summoned to the palace said it plainly: “You should resign.” The man who had ruled for thirty-two years replied, “I’m not going to argue with that.” On the morning of May 21, in under ten minutes, he was gone.

In Seoul, in June, an insurance company lays off seventy-six salesmen. One of them buys a hundred hats.

The bill

Camdessus had promised “a relatively short, but sharp, weakening”. What came was the deepest slump in the region’s modern history. In 2000 the Fund admitted its fiscal targets “were too tight.” Indonesia’s real manufacturing wage fell by about a quarter in 1998.

Now the two ledgers, closed. Protected: the funds, out by August. The trading desks that “made money out of volatility.” The banks that arranged the loans, with their record fees. Korea’s creditors, with a state promise at a higher rate. Indonesia’s best-connected families, relocating to Singapore and Hong Kong while their bank-loan settlements went mostly unhonored; a US embassy cable later put the unpaid share of those loans at “approximately one-third.” Paid:Indonesian taxpayers, who absorbed a bank cleanup costing more than half a year’s national output in public bonds. 3.5 million people in gold. Some twenty-five hundred more Korean suicides than the year before. And four people you met in the last ordinary summer.

Youn Sung Mook, the X-ray technician, sleeping in the lobby of his closed hospital because he could not afford the fare home: “Every aspect of my life has changed.” Kim Myung Yun, 39, ran a sales team at Korea Life until the end of June, when he was laid off with seventy-five others. He had severance, and a plan. He and his brother-in-law bought a hundred hats and sold them from a car, pinned to green mesh with clothespins, on a Seoul pavement. A hat cost four dollars and made a dollar and a half, to be split two ways. On the day the Post’s reporter stood with him, he worked fourteen hours and sold two. After train fare he had cleared about twenty-five cents. Debt had eaten most of the severance and half of a $750 monthly unemployment payment that would stop in December. “I can’t make anything work out.” Zen Zainudin, the singer, on a blue metal bench outside the employment office: “All I can do is try as hard as I can and pray.” Warih Wijayanti, the architect, teaching primary school for $40 a month, down from $135.

None of them mispriced a currency. None of them chose a bank’s funding strategy or earned a fee for arranging anyone’s borrowing. None of them was at Camp David, or in the restricted session, or on the Fed’s line, or among the investment houses with a stake. All of them paid. What the file cannot say is whether any rescue could have spared them. What it says is who was asked to bear which risk. That much is not opinion. It is in the file.

The two ledgers, July 1997 to May 1998: who was protected, and who paid. Every line is sourced in the text above.

The next payment date

This is history, but the bills are not. On September 1, 2026, Korea’s cabinet approved a budget proposal that, in the government’s own words, “completes the repayment of the public funds injected to respond to the 1998 foreign-exchange crisis.” The Financial Services Commission’s count: 168.7 trillion won went in from November 1997. By June 2026, 72.9 percent had come back. The last instalment falls in 2027, thirty years on. In Thailand, the 0.46 percent levy on bank deposits that services the 1997 rescue debt still had some 460 billion baht of principal to clear at the end of August 2026. In Indonesia, the Suharto-era foundation ordered by the Supreme Court in 2015 to repay some Rp 4.4 trillion still carried some $307 million of it in the Finance Ministry’s 2025 accounts, as read for this essay. Three countries, one direction of travel.

To be exact about what was guaranteed and what was paid. The guarantee on the $21.8 billion of rolled-over bank loans was never called; the Korean banks paid the state a fee for it, and by the Fund’s own account the loans were repaid in full by April 2001. The public money was something else: 168.7 trillion won put into banks and the deposit insurer to absorb the losses of the collapse, of which 72.9 percent has come back, and whose last instalment is the one in the 2027 budget. The guarantee protected the lenders from a risk that did not arrive. The public funds paid for the losses that did.

And the doors are the same doors.

The Camp David door. On March 19, 2020, the Federal Reserve opened a $60 billion swap line to the Bank of Korea, one of nine it opened that day as the pandemic hit, as it had first done in the crisis of 2008. It is a different instrument from the loan Korea begged for in 1997, and a different decade. But the substance is the one Kim Young-sam could not get in nineteen minutes: dollars from Washington without a program. The rulebook is unchanged. Five foreign central banks have standing lines; the rest are let in when Washington decides; everyone else posts US Treasury collateral.

The Treasury door. In January 1998 a White House aide wrote that he was “trying to fudge some on the ESF,” the Exchange Stabilization Fund. In October 2025 that fund wrote Argentina a $20 billion swap and bought pesos in the market. Two years earlier, Argentina had paid the IMF in Chinese yuan drawn from a swap line with Beijing. The doctor has a rival now. Beijing’s swap lines have provided more than $170 billion in rescue money, a rival conceived in the rooms above, as Part 3 will show. Which door a finance minister knocks on is now part of every credit decision, and every alliance.

The street. In August 2025 Jakarta’s students were back at the parliament, this time over a housing allowance for legislators of Rp 50 million a month, about ten times the capital’s minimum wage; a motorcycle-taxi driver died under a police vehicle. The president they were protesting to is Prabowo Subianto: Suharto’s son-in-law, the special-forces commander who told an American assistant secretary in November 1997, in a cable marked SECRET, that “the Soeharto era will soon end.” He was dismissed from the army in 1998 amid allegations over the kidnapping of activists, which he has always denied; he was never tried; he was sworn in as president in October 2024. In Seoul, in February 2025, the mint suspended gold bar sales and the vending machines sold out. In 1998 the queue was to hand gold in. In 2025 it was to buy.

Part 1 opened with the AI build-out and the bonds that pay for it. The three questions of 1997 are the same three: who holds the paper, when does it roll, and who is the lender of last resort when it stops rolling. Add the fourth, which is this essay: what do they want in return.

Five things 1997 teaches you to watch, as observation, not advice. Reserves net of forwards, never the headline. The rollover schedule, not the debt total. The interest rate, first casualty of a currency peg. Who the lender of last resort is, and what they want. And every scene happens twice: the file is the information.

You have now read the file, minus the four memos held back. A worker’s overdue wage has a payment date too. When the next rescue is announced, skip the speeches and read the baseline projection. Who pays is usually already in it. Envisaged, in the original.

Part 3: The Counterattack. The territory that refused: Hong Kong’s ten days in August, the $15 billion ambush at the futures expiry, the Malaysian everyone laughed at closing his borders. And an IMF economist’s private email of January 30, 1998, preferring “capital controls” that “would at least impose some costs on foreigners.” Who got paid, and who paid.

The file — read it yourself

The Vancouver memoranda, Nov 24, 1997, SECRET, released in full: Clinton–Suharto and Clinton with the ASEAN leaders; the Deputy Secretary’s Jakarta cables, Nov 7, 1997: Talbott–Suharto (CONFIDENTIAL; Suharto asks for rules and a tax on currency trading, Talbott asks him to stop an Indonesian oil deal with Iran) and the American business lunch; Talbott’s Nov 12 meeting in Tokyo, where the US position on an Asian fund is set down — “concerned with anything outside the IMF, even with ‘linkage’ to IMF conditionality.” All from State Department litigation release F-2017-13804. The Defense Intelligence Agency report on Trisakti, May 13, 1998, via the National Security Archive. Australia’s National Security Committee minute, March 17, 1998, NAA A14370, JH1998/85. The present day: Federal Reserve press release, March 19, 2020; Congressional Research Service, U.S. Financial Support to Argentina (R48780); Al Jazeera, July 17, 2023, and Yale SOM’s account of the yuan payments; IMF, Review of Charges and the Surcharge Policy, October 2024; JAKARTA 006622, Nov 10, 1997, SECRET, released in full, via the National Security Archive; CNN, October 20, 2024; University of Melbourne, Indonesia at Melbourne, on the August 2025 protests; CNBC, March 6, 2025, and the Korea Times, February 2025, on the gold shortage. The Christmas Eve meeting: William McDonough’s own account (PBS interview, April 17, 2001), the Federal Reserve’s history essay (Carson and Clark, 2013), and Treasury statement RR-2131of December 24, 1997; the bankers present are not named in those records. The CIA’s National Intelligence Daily, Special Analyses, “How a Coping Scenario Could Unravel,” Nov 19, 1997 (PASS NID 97-0269CX; Top Secret; released Feb 4, 2009). The presidential calls: Nov 27, 1997, Camp David (State Department release, 2019; also in Clinton Library MDR 2016-1056-M); Nov 28, Clinton–Hashimoto (cable STATE 227100); Dec 3 (cable STATE 230759; also Library item 118553); Dec 12, Dec 16, Dec 18 and Jan 12, 1998 (Clinton Library). The Talbott–Yoo memorandum, Dec 11; the Talbott–Mar’ie cable, Nov 6; the WTO talking-points cable, Dec 12 (STATE 232229). The IMF’s open-session Korea minutes of Dec 9 and Dec 19; the Korea staff paper, EBS/97/222; the first review, EBS/97/237; the Asia department’s Korea crisis file for December 1997 (Lissakers Dec 5, Coorey and Fisher Dec 8, Leddy Dec 9; declassified 2019) and its Indonesia filefor January 1998; BUFF/ED/98/4, the Indonesian statement of Jan 8, 1998; Australia’s Aug 18 statement on Thailand; the Dec 30 allowance schedule. The World Bank Group Archives folders 30489567 (Korea pricing, Feb 1998) and 30488850 (Nov 1997), and Report P-7225-KO. The Bank of England Court minutes for Sept–Dec 1997 and Jan–May 1998. The White House communications file (MDR 2014-0154-M). The Federal Reserve’s Dec 16, 1997 transcript. The National Security Archive’s Indonesia collection.

Standing on shoulders: Blustein and Chandler’s Christmas-rescue reconstruction; Richburg’s “Seven Days in May”; the New York Times’ “Global Contagion”; Mary Jordan’s and Kevin Sullivan’s street reporting.

A note on the records

What this part adds. As far as I can find, these records have not been quoted before: the Camp David transcript of November 27, 1997; the Vancouver memoranda of November 24; the Deputy Secretary’s Jakarta cables of November 1997; the December memos in the Fund’s Asia department file; the CIA’s daily brief of November 19; Australia’s National Security Committee minute of March 17, 1998. The connections are mine, and readers should weigh them as such: the finance minister’s “no need” and the CIA’s “may be insufficient” on the same Wednesday; a White House aide’s “leverage” beside Canberra’s “social stability”; the fax of December 8 beside the room on Christmas Eve. The reserves table is a projection under stated assumptions, not a record of what happened.

The presidential memoranda were prepared by White House notetakers; the IMF minutes are summary records; quotations reproduce the records’ wording, checked against the scans; cable text, printed in capitals in the originals, is set here in ordinary type. CONFIDENTIAL, SECRET and STRICTLY CONFIDENTIAL are the markings on the originals; NOT FOR PUBLIC USE is the IMF’s filing stamp, not a security classification. The Nov 28 call survives as a State Department cable of the White House text, dated Dec 4; it prints no call time. The Dec 11 Yoo memorandum gives Korea’s usable reserve as ten billion dollars in one paragraph and five billion in another. The IMF staff reserve table is a projection under stated assumptions (30 percent rollover, $1 billion a day falling due), not a record of what happened; the “substitution of sovereign debt for private debt” memo frames Korea’s problem as liquidity, not overspending, and argues for financing it. The two 2008 embassy cables are leaked texts published by WikiLeaks; the convictions they describe are confirmed by the Indonesian court reporting linked in the text (the embassy’s bribe figure, $600,000, differs from the court’s $660,000). No individual is described here as guilty of anything a court did not find. The Korean budget line is the government’s own wording, translated; the budget was a proposal at the time of writing. The Supersemar balance for 2025 comes from the Finance Ministry’s audited Treasurer’s statements as read by the research lane, not yet re-opened by the author.

Notes and sources

1. Korea’s creditor deal: 134 banks, $21.8 billion, one- to three-year loans, 2.25–2.75 points over six-month LIBOR, Korean government guarantee — per the IMF’s Korean Crisis and Recovery (ch. 2), the Korea Development Bank’s 2004 SEC prospectus (pp. 76–77) and IMF Country Report 00/11 (p. 31). The guarantee was never exercised; by the IMF volume’s account (ch. 12) the exchanged loans were repaid in full, without default, by April 2001; the guarantee fee was paid by the Korean borrowing banks to the Korean state, not by the foreign creditors. The guarantee fee (about $150 million, 0.2–1.5 percent a year, paid by the banks to the state) is Kim and Byeon’s account in the same IMF volume, written from ministry records. Wall Street bonuses: $11.2 billion for 1997 — New York State Comptroller series. Jaguar’s 72 percent is gross and whole-fund; no full-year figure is an Asia profit. Morgan Stanley Dean Witter and Chase figures are the firms’ own global disclosures (1998 Annual Report, p. 30; 8-K of Jan 21, 1998); neither is an Asia profit figure.

2. The open-session minutes: the “$65 billion plus about $50 billion” exchange (with staff’s caution about possible double counting) and the interest-rate ceiling are on printed pages 3–5 of the Dec 9 minutes. The restricted-session debate of Dec 4 remains unrecorded; EBM/97/115 (Dec 3) contains no Korea discussion.

3. The Korea crisis file (IMF Archives, Asian and Pacific Department Director’s Office Files – Asian Crisis, Box A8-047, file 6): Lissakers to Kwon, Dec 5 (PDF p. 28); Coorey to Neiss, Dec 8 (pp. 135–136); Fisher to Boorman, Dec 8, with table (pp. 138–140); Leddy, Dec 9 (pp. 110–116); all STRICTLY CONFIDENTIAL or SECRET, stamped DECLASSIFIED. Four further memos in the file — “Korean Loan” (Fischer, Dec 2), “Meeting with Executive Directors” (Dec 9), “Indonesia and Korea” (Dec 9) and “Korea – Contingency Planning” (Leddy, Dec 9, SECRET) — remain withdrawn; their withdrawal notices are in the file.

4. “Virtually paralyzed”: Korea first review, EBS/97/237, printed pages 10–12; the staff continued to support the program.

5. Dates the record disputes: Taiwan’s move (mid-October); Korea’s request (announced the night of Nov 21; the IMF’s chronology says the 20th); the won’s limit-down (Nov 19, per the Post; the Bank stopped defending it on the 17th); Trisakti’s first-day toll (five reported, four established); several Washington Post archive pages display one date while their URLs carry the next day.

6. The suicide figures: roughly 6,000 deaths in 1997 to 8,600 in 1998 (≈42 percent) and a 45 percent male-rate increase (Chang et al., 2009; Korean statistics reporting). Unemployment: 2.6 percent (1997) to 7.0 (1998), annual averages, Korea’s SEC filing. Indonesian real manufacturing wages: about −25 percent in 1998, ILO Employment Report 3. These figures have separate populations and measures; their juxtaposition establishes different outcomes, not a traced transfer between individuals. The gold campaign: the Deseret/Reuters report is from its first week (17 tons); the final figures (about 227 tons, some 3.5 million participants) are the widely reported totals.

7. Indonesia’s cleanup cost (56.8 percent of GDP, gross) is the Laeven–Valencia IMF database; the settlements’ performance is the IMF’s 2002 review; elite relocation is contemporaneous reporting and does not establish individual transfers.

8. Who would write the rules: the WTO cable (STATE 232229, Dec 12, 1997, CONFIDENTIAL, declassified; cleared “TREASURY:LSUMMERS”; it says the US asked Korea only to bind “that which it has already agreed to do”); the Blinken email (Dec 31, 1997, unclassified) and Jan 6, 1998 memorandum (CONFIDENTIAL); the foreign-ownership ceilings in EBS/97/222. The World Bank’s pricing: folder 30489567 (”Options for FY98 Net Income,” Feb 24–27, 1998) — Wheeler to Wolfensohn, Feb 26; Won to Aiyer, Feb 25; the Feb 17 pricing note (Directors’ proposals had ranged to “LIBOR+500bps”); Report P-7225-KO (Mar 19, 1998). The Bank’s stated reason for the pricing was its own risk-bearing capital, not profit; the words about earnings are its own, describing the December loan. Severino’s Nov 24 letter is in folder 30488850; Fischer’s Jan 7 memo is cited from the Clinton Library release.

9. Participant-recalled dialogue (the Madjid–Suharto exchange) is printed as attributed recollection. The Clinton–Suharto lines are from the declassified telephone records.

10. The bills today: FSC release of July 21, 2026, and the Ministry of Planning and Budget’s FY2027 proposal (Sept 1, 2026); Bank of Thailand FIDF status table, Aug 31, 2026, corroborated by the Public Debt Management Office and a Bangkok Post report of 478.177 billion baht at March 31, 2026 (an 18 billion baht transfer intervened); The Nation, Nov 6, 2025, for the levy and the six-versus-ten-year expectation; detik.com and ANTARA (2018) for Supersemar; LKBUN 2025, pp. 165 and 489–490, for the 2025 balance. The live wires are reported, dated figures from the linked sources and illustrate funding arrangements without assessing the countries involved as of publication. Nothing here is a forecast or a recommendation.

Hong Kong, handover night, July 1, 1997. The world watched the fireworks. In Bangkok, the central bank had $2.8 billion left. The next morning, it let go. (Photo: Ferdi Hartung / Süddeutsche Zeitung Photo / Alamy)

This series tells the story of the Asian Financial Crisis of 1997–98 for readers who weren’t there. It is built on primary documents — central bank inquiries, declassified transcripts, the IMF’s own records — and every load-bearing claim is linked to its source. Part 1 is the way up. Part 2 is the collapse. Part 3 is the bill, the fight-back, and what it means now.

On June 8, 2026, a company that has never earned an annual profit filed to go public. OpenAI was reported earlier this year to be losing about $14 billion a year. Its last funding round valued it at a reported $852 billion. Its signed commitments for computing power total roughly $1.15 trillion, spread over the next decade in deals of varying legal firmness.

That is the faith-priced layer. Beneath it, the most profitable companies in the history of money have started borrowing. Alphabet just posted the first negative free-cash-flow quarter of its public life, and its long-term debt roughly doubled in six months. The big AI builders are guiding toward $660 to $690 billion of capital spending this year, estimates rising with each earnings call, and a growing share of it is financed in the bond market.

Nobody doubts. That is what a boom is: a machine for not doubting.

Some of you are already objecting, and you should. Today’s giants earn real profits. They borrow in a currency their own country prints. The technology works, the demand is real, and this is nothing like some emerging-market blowup from before you were born. Good. Hold that objection all the way down this story — because it is the story of a place where every one of those reassuring sentences was also true. Asia’s growth was real. The factories were real, the exports were real, the rising wages were real. The World Bank called it a miracle in an official publication. The collapse happened anyway. Not because the miracle was fake, but because of the plumbing under it: who had lent what, to whom, due when, in whose money.

A valuation is an opinion. Debt is a schedule. This is a story about schedules.

You have seen this film before. You just weren’t born in time for the first screening.

Bangkok, 1994.

The miracle was real

Start by believing, because everyone did, and they were not fools.

For a generation, the economies of East and Southeast Asia grew close to seven percent a year — doubling in size every decade. Hundreds of millions of people left poverty. Villages got electricity, then schools, then daughters in universities. In September 1993 the World Bank made it official, publishing a study called The East Asian Miracle and telling the rest of the developing world: this can be repeated.

The world had a nickname for these economies — the Asian Tigers, with a litter of “tiger cubs” coming up behind — and the region believed its own press, with reason. Malaysia published a national plan to be a fully developed country by 2020 and started building the tallest towers on earth. Leaders lectured the West that “Asian values” — discipline, savings, family — were the deeper engine of the boom. The Petronas Towers were finished in 1998. By then, the crisis was raging beneath them.

This is the moral pivot of the whole story. The boom was not a scam. Even the document Indonesia later signed in surrender — the January 1998 Letter of Intent, signed under the crossed arms of a photograph you will meet in Part 2 — opens by conceding it: decades of growth, poverty “dramatically” lowered, budgets balanced. Which is exactly why what came next was not a correction. It was a taking. And the weak point was never in the growth statistics.

It was in the plumbing. So let’s go look at the plumbing.

The promise

What a currency peg is. Thailand’s baht was held near 25 to the US dollar, year after year. Formally it was pegged to a secret basket of currencies — but the basket was mostly dollar, and the central bank defended a daily dollar price. In effect: a dollar promise. A peg is a promise. And a promise kept long enough stops being a price and becomes a fact of nature. Businesses plan around it. Banks lend around it. Nobody insures against it — who insures against the sunrise? In financial language, nobody hedges. A Thai company could borrow dollars in New York at 6 percent, swap them into baht earning 12 at home, and pocket the difference, with the peg guaranteeing the exchange back. Free money, as long as the promise held. Traders call this a carry trade. Keep the name. It comes back later in this series wearing yen — and it has visited your own decade wearing yen too. A whole country, in effect, sold insurance against its own devaluation. And spent the premiums.

The peg made Thailand feel like a sure thing. The only question left was how much money a sure thing can attract. The answer turned out to be: all of it.

But the foreign money did not just wander in. It was invited — by name, by law, at a discount.

Bangkok opens a window

The invitation has a paper trail. In April 1990 the governor of the Bank of Thailand endorsed a plan — quoted later by the official inquiry into the disaster — to make Bangkok a regional financial hub, a little Singapore. Competitiveness. Prestige. Stability. In March 1993 the plan got its machine: the Bangkok International Banking Facility, some four dozen licensed banking units allowed to raise money abroad and lend it onward.

On the brochure, this was an export business: raise dollars offshore, lend them offshore. In the fine print sat a second channel — raise dollars offshore and lend them into Thailand — and one number that decided everything. BIBF business was taxed at 10 percent, against the standard corporate rate of 30. The Thai state taxed the business of piping in foreign debt at one-third the rate it taxed everything else.

You can guess the rest. A fast-growing country. Twelve percent interest at home, six percent dollars abroad. A peg guaranteeing the round trip. A tax discount on the pipe. Foreign lending into Thailand through the facility went from 195 billion baht at the end of 1993 to 807 billion by the end of 1996 — and most of it was short-term, loans that had to be renewed every few months. A later Bank of Thailand research paper describes the inflows rising “at an unprecedented rate.”

Nobody decided Thailand should owe the world tens of billions of dollars due within the year. They decided to build a hub, cut a tax, keep a promise. The debt assembled itself out of incentives, one rational transaction at a time. A catastrophe needs less villainy than you think.

Seoul’s beautiful, backwards caution

Korea built the same bomb out of the opposite instinct.

When Seoul began opening its financial account in 1993, its officials were afraid of foreign money. So they liberalized the way that felt safest: they eased banks’ short-term foreign borrowing while keeping long-term borrowing restricted, because short-term openings felt reversible. You can always let a three-month loan run off. Long-term money was the marriage. Short-term was just dating.

Two accelerants turned caution into catastrophe. The first had been running since 1988: under the Basel banking rules, a loan to a bank outside the rich-country club carried one-fifth the capital charge if it ran under a year — the world’s regulators were, in effect, paying banks to lend to Asia short. The second arrived when Korea joined that club, the OECD, in 1996: now lending to Korean banks was cheap at any maturity, and the money came faster still. Most of it flowed through lightly supervised merchant banks, the weakest door in the building. Korea’s short-term external debt went from $40 billion in 1993 to $98 billion by September 1997 — more than half the country’s entire external debt, re-borrowed every few months from lenders who owed Korea nothing.

Hold the shape of it: what feels prudent to each decision-maker can be fatal in aggregate. Reversible for the lender means runnable for the borrower. Korea had cautiously, deliberately, built the world’s largest bank run in waiting — and a clause in a Swiss rulebook helped pay for it.

The only show left in town

Now cross to the other side of the pipe, because this is where this section’s name comes from — a real headline, from the boom itself.

February 1995. Mexico has just blown up, and for a moment investors are afraid of everything called “emerging.” Euromoney — the trade magazine international bankers wrote for each other — runs its Asia coverage under a headline of perfect period honesty: “The only show left in town.” The bankers quoted are nearly unanimous: the sell-off is irrational, Asia is where the decade will be earned. And a Citicorp executive in Hong Kong hands history a line: “We are sitting on the tip of a volcano which will erupt.” He adds: “The only question is when.”

He meant demand. He meant the money still trying to get in. A banker, two years before the eruption, chose the volcano as his metaphor for opportunity.

Borrowing short in someone else’s money. Every loan has two clocks: how long the money is promised, and whose currency it must be repaid in. Borrow long in your own currency, and trouble means renegotiation. Borrow short in someone else’s, and trouble means one phone call: repay me in dollars, this week. Whoever lends short holds the exit. Remember this box. It is the whole series in four sentences.

The money came fast and it came big. Syndicated lending to developing countries nearly doubled in two years — $58 billion in 1993 to $112 billion in 1995 — with East Asia driving 44 percent of the increase. By mid-1997, foreign banks were owed about $274 billion by Thailand, Indonesia, Korea, Malaysia and the Philippines. Nearly two-thirds of it came due within a year.

The record also corrects the folk memory here. The story is usually told about Japanese banks, and in Thailand and Indonesia they did lead. But it was European banks that lent the most across the region — the largest creditor bloc in Korea, Malaysia and the Philippines, their share still climbing as the bubble peaked. Frankfurt, Paris and London were at this party at least as hard as Tokyo. New York arranged the music.

And the economics of the lending were strange. By 1996, blue-chip Asian borrowers were paying under 0.2 percent over the banks’ own cost of funds. At those margins the loan itself earned almost nothing. So why fight to make it? Because the loan was the ticket. It bought the relationship, and the relationship paid in everything else: underwriting mandates, advisory work, foreign exchange, the league table. No audited total of those fees survives — what survives is the incentive, and the incentive was to keep the pipe full.

Be fair about what this was, and wasn’t. It was not a heist with an exit plan; many of these banks were later mauled by the collapse they helped build. It was something more ordinary and harder to fix. Every quarter you kept lending, you got paid. Every quarter you stopped, someone else did. So ask the question this essay keeps asking: when the loans go bad — and they will — who has already been paid?

The insiders

Foreign lenders need local borrowers, and the local systems were built for taking.

Jakarta, 1992: a bank called Summa collapses, and the autopsy finds roughly 70 percent of its bad property loans went to companies in its own group. It is a perfect early warning. Deregulation had spawned dozens of private banks that worked, in practice, as wallets for the families that owned them. Lending limits existed on paper. Enforcement did not. Nothing changed. Above the banks sat the concession economy: a marketing board every clove farmer had to pay, a “National Car” with special tax and credit privileges, a plywood cartel every exporter served. In Part 2, the IMF will demand all three by name, in writing — and it can name them because everyone in Jakarta always could.

Seoul: the chaebol, the family industrial groups, run on debt at ratios no Western board would sign, rolled through those lightly watched merchant banks. Bangkok: a boom-fed crowd of finance companies shovels credit into towers and golf courses. The flagship scandal is already public — the Bangkok Bank of Commerce, bad loans near 40 percent of assets by 1993, which Thailand’s official inquiry found had made takeover loans for insiders. When parliament finally said so aloud in May 1996, depositors ran and the central bank seized it.

Two things are true at once, and this series will hold both to the end. The insider economy was real; the cronyism was not a Western slander. And the foreign money saw all of it, priced none of it, and kept lending — because the fees were good and the exit was short. The corruption did not hide from the lenders. It was the collateral they accepted.

One more thing before we leave this room. Fix the shape of what you just saw. Capital and revenue circulating inside a family of connected companies. The bank owned by the borrower. The group guaranteeing itself. Every single transaction defensible; the total risk invisible from outside. You will need that shape at the end of this essay, closer to home than you expect.

Everyone knew — and nobody moved

Not everyone knew, to be honest. The market priced in almost nothing; Thailand borrowed at rich-country spreads into the spring of 1997. The warnings lived in documents, not in prices. But the documents exist, and they are damning.

Krugman’s famous 1994 cold-water essay had said only that the growth would slow — sweat, not magic — never that it would crash. The sharper warnings were closer to the plumbing. The IMF’s board reviewed Thailand in July 1996 and mostly heard a success story — the American director called it “a rather extraordinary success story” — even as Camdessus wrote privately to Bangkok days later urging more exchange-rate flexibility. Moody’s put Thailand on review in 1996 and downgraded its short-term debt that September; money noticeably left. And inside Bankers Trust, an analyst named Aaron Henderson filed research whose titles alone tell the story — “Thai Bills of Exchange: Storm Warning” (August 1996), then “The Thai Finance Sector: The Storm Has Begun” (October). The reports themselves have not survived online; his titles live on in a scholar’s footnote. Note where he sat when he wrote them: inside one of the lending banks — an institution that, months later, helped lead the creditor talks on Korea’s debts. The system’s own analysts saw the storm. The system’s own incentives filed the memo.

Why do warnings fail? Not stupidity. Payroll. A fund manager who exits a boom two years early is fired two years before being right. A bank that stops lending hands the league table to the bank across the street. A finance minister who devalues early owns the pain; one who waits can blame speculators. Everyone is paid to dance until the music stops. The chief executive of Citigroup said it best, a decade later, months before his bank began writing down billions: “As long as the music is playing, you’ve got to get up and dance. We’re still dancing.”

That is where this series got its name.

And notice who did act on the warnings. Not the regulators. Not the lenders. In New York, a small tribe of fund managers read the same trade data and the same reserve tables you have just read — and reached for the phone. George Soros put it plainly afterward: “The trouble could be seen six to nine months before the decline actually occurred.” The people who believed the warnings were the ones paid to bet on them.

They had seen the vault. We are getting to it. But first, meet four people who had never heard of any of this.

Four lives, before

In Seoul, Kim Myung Yun runs a sales team at an insurance company — salaryman-solid, a man whose business card does the introductions. Across town, Youn Sung Mook, thirty-two, works as a hospital X-ray technician. A licensed man in a white coat; the kind of job a Korean mother brags about. In Jakarta, Zen Zainudin sings in a karaoke club — night work in a city with money to spend on song — and Warih Wijayanti, an architect, draws the skyline the foreign money is paying to raise.

No syndication desk knows their names. Their pay comes in won and rupiah — the promise currencies. Between these four people and the most crowded exit in financial history stands, though no one has told them, nothing at all.

They come back in Part 2. Their jobs don’t.

The vault

Now the concealment, month by month. Watch the dates.

Through 1996. Thai exports stall. The current account deficit reaches nearly 8 percent of GDP — the country spending far more than it earns, the gap financed by exactly the short-term inflows you now understand. Towers stand empty. Finance-company loans rot quietly.

February 5, 1997. A property developer called Somprasong Land misses an interest payment on its foreign debt — the first Thai corporate default of the era. A small sound, like ice cracking somewhere upriver.

The same weeks, in private. The IMF is blunter than anyone will know for years. On January 31, 1997, Michel Camdessus — who had already made a hastily arranged stop in Bangkok in December — writes privately to Thailand’s finance minister: move “quickly and decisively.” A week later his deputy, Stanley Fischer, writes again, harder: defending the peg means punishing interest rates and — his words — “the risk of a rapid rundown in reserves.”

March 1997, in a different room. The same Camdessus tells his own executive board — in minutes quoted only when the Fund’s historian published them fifteen years later — that there is “certainly no reason for panic in Thailand.” And to the press he goes further: the Thais, he tells reporters, are doing “exactly what you must do” to avoid a Mexico-style crisis; “I don’t see any reason for this crisis to develop further.”

Private alarm to Bangkok. Private calm to the board. Public reassurance to the press. Hold that triple gap for the rest of the series, whenever anyone tells you the officials were simply blindsided.

Gross versus usable reserves. A central bank’s “reserves” — its war chest of dollars — is a headline number. But a central bank can also sell dollars forward: promise to deliver them next month, at a fixed price, without touching today’s headline. Defend a peg through forwards, and the war chest looks untouched while being quietly committed. The forwards settle over months, so the netting is not one-for-one on any given day. But at the moment of attack, promised dollars are not spendable dollars. Thailand’s headline said roughly $38 billion. The promises were becoming enormous.

Spring 1997. As pressure builds, the Bank of Thailand defends the peg more and more through the forward market — invisibly. The scale is hidden from the public, from parliament, and in any usable form even from the IMF, whose staff is reduced to inferring intervention from reserves that stay implausibly still. Inside the bank, the paper trail accumulates, waiting for the inquiry that will find it: intervention memos of May 9, 12, 14, 15 — each annotated in the governor’s own hand with the same sentence: “The Finance Minister has been duly kept informed.”

Everyone who mattered was, in writing, informed. No one who mattered was accountable. You will meet this sentence again, in other countries and other decades.

End of June 1997. Headline reserves: about $31–32 billion. Committed forward: about $29 billion. Freely usable: around $2.8 billion. A $38 billion fortress had become a stage set. Almost no one outside the bank knew.

Three days before

The professionals had smelled the gap between the promise and the arithmetic months earlier. From February through May 1997, macro hedge funds built short positions against the baht — about $7 billion of what became a $29 billion book, by the estimates participants later gave the IMF. Set the folklore next to the number. The funds were one seller among four, alongside offshore banks, foreign banks inside Thailand, and — in time — Thai companies scrambling to cover unhedged dollar debts. The wolf pack was real. It hunted in a forest the shepherds had already sold.

May 14. The attack crests. The Bank of Thailand commits more than $6 billion in a single day defending the promise, mostly through forwards.

May 15–16. The defenders counterattack, and brilliantly — a squeeze that cuts speculators off from baht and briefly drives their borrowing cost past 1,000 percent annualized. At that rate, holding a short position costs nearly three percent of it per day. The traders howl. Bangkok declares victory. The victory has consumed, in effect, the last of the war chest.

May 22. The IMF’s number two, Stanley Fischer, slips into Bangkok — driven through the city, he later recalled, in a van with darkened windows. The prime minister refuses to see him. The warning is hand-delivered anyway.

June 19. The finance minister — the man duly kept informed — resigns.

Sunday night, June 29, 1997. The Prime Minister of Thailand, Chavalit Yongchaiyudh, goes on live national television to steady his country, and rules out devaluation: “To devalue is easy, but it could cause bad damage and affect all Thais.”

You know what he does not say. Usable reserves: $2.8 billion. In Seoul and Jakarta, for the four people you met, it is one more ordinary Sunday. Nobody has told them anything.

Three days later, on the morning of Wednesday, July 2, a brief official statement changed the price of everything.

Part 2: When It Stopped — the morning statement and what it did to the price of everything; the wolf pack’s best day; sixteen banks closed in a single Jakarta announcement; the secret Federal Reserve transcript that records the word [Laughter] eight days before a nation nearly runs out of dollars; and the photograph — arms folded — that helped end a 32-year reign.

The mirror, before you go

This is a history, not a prophecy. But you did not read it for the nineties.

Global debt ended 2025 at a record $348 trillion — about three times world output — with $29 trillion added in one year, though as a share of output the ratio edged down. Emerging economies face a record $9 trillion coming due in 2026. A private-credit industry the FSB sizes at $1.5 to $2 trillion now does much of what banks once did; the FSB’s own report notes gaps in the data and says the sector “has not been tested during a severe economic downturn.” Our era’s quiet BIBF. And the AI build-out has the technology sector selling bonds at a pace that invites comparison to the railway age — issuance that, the OECD notes, markets have so far absorbed calmly. So far.

And this very month, the US Treasury began what its own secretary calls a “Treasury Twist”: doubling its buybacks of long-term bonds and paying for them with newly sold short-term bills — deliberately shortening the schedule of the world’s anchor debt to hold long-term rates down. Analysts call the size more signal than substance, so far. But notice what even the strongest borrower on earth does when the long end gets expensive: it moves its debt closer to the door.

Name the differences honestly, because they are real. America borrows in a currency it prints. There is no peg to break, and no 1997 is possible for the giants at the center. But 1997 was never about the center. Thailand’s blue chips were not the fault line; the leveraged, faith-priced layer around them was. So look where that layer sits today: the specialist AI-cloud firms borrowing at high yield against chips that age in four years; the private-credit funds; and, offshore, everyone on earth who borrowed dollars to join a boom priced in someone else’s currency. That was Thailand’s seat at the table.

Now recall the shape you were asked to fix in Jakarta — money circulating inside a family of connected companies, every deal real, the total unknowable. It has been rebuilt, at scale, in the middle of the AI boom: a reported $750 billion web of interlocking deals in which Nvidia invests in OpenAI, which pays Oracle and CoreWeave for computing, which spend the money on Nvidia’s chips — the vendor holding equity in its own customers, revenue and capital chasing each other around a loop that makes true end-demand hard to read from outside. The chaebol guaranteed each other’s debts. The ecosystem now finances each other’s purchases. It is not fraud, and nobody is hiding it — neither were the Jakarta bank-wallets, and that is precisely the point. Connected finance fails legally, in public, all at once. When the chief executive of one AI-cloud firm called the claim that Nvidia’s stake props up his company “ridiculous” — the stake is a few hundred million dollars against the many billions his firm has raised — he may well be right about his company. The point was never any one company. The point is the shape.

To be fair to the present: the Asians learned. The countries in this story armored themselves in reserves precisely so it could never happen to them again — that armor, and who still has none, is Part 3. The fragility didn’t vanish. It moved — and it always moves to wherever the schedule is shortest.

So carry one sentence out of 1994 into your own decade: whoever lends short holds the exit. The music is excellent. It usually is. The only question that ever matters is the one nobody on a dance floor asks: who is standing nearest the door — and in whose currency is the bill?

The file — read it yourself

This series is built so you never have to take our word. The core documents behind Part 1, in the original: Thailand’s official inquiry into the disaster, the Nukul Commission report (as published in translation by The Nation); Indonesia’s January 15, 1998 Letter of Intent, with the clove board, the National Car, and the cartels named in the text; the Federal Reserve’s secret December 16, 1997 transcript, released years later; the declassified White House–Suharto records at the National Security Archive; and the BIS’s 1998 accounting of who had lent what.

And know that one drawer is still closed: a UK Treasury file on Hong Kong’s financial markets and exchange rate for 1996–97 — reference T 559/42 — is marked “Closed Or Retained Document,” held back by the department under section 3(4), a quarter-century later. Investigations end at locked drawers. Series don’t. We will be back at that drawer in Part 3.

Standing on shoulders. A deep journalistic record of this crisis exists, and it is worth your time: Paul Blustein’s The Chastening (2001), reported from inside the IMF’s missions through more than a hundred interviews; PBS Frontline’s The Crash (1999), whose interviews with the principals are online in full; PBS’s Commanding Heights oral histories; and The Washington Post’s 1998 reconstructions, several of which supplied the human threads in this series. Where this series adds anything to their work, it is the paper: boardroom minutes, transcripts, and cables that were still sealed when those accounts were written.

Notes and sources

Method. Researched through two independent AI research lanes, cross-checked against each other, then stress-tested by an adversarial fact-check against primary and contemporaneous sources. Load-bearing historical claims are verified against primary or contemporaneous documents and linked at first use. Figures about 2026 are reported estimates from the linked outlets, dated as of late August 2026, and re-audited before each part. Where the record is contested or incomplete, the text says so.

1. 2026 figures. OpenAI’s losses, valuation, and compute commitments are reported figures, not audited disclosures; its IPO filing is confidential and its terms are not public. Capex guidance ranges are from company earnings calls as compiled in the linked analyses and have generally been revised upward during 2026. The IIF debt figures are from its February 2026 Global Debt Monitor. The Treasury buyback expansion is as announced August 19, 2026.

2. Reserve figures distinguish gross reserves, forward commitments, and usable reserves. The IMF’s contemporaneous analysis cautioned that netting forwards one-for-one against reserves overstates the daily drain, since forwards settle over months; the sidebar says so. The $2.8B usable figure at the float is the Nukul Commission’s; end-June figures from IMF WP/99/138.

3. The hedge-fund $7 billion is a participant estimate assembled by the IMF in its 1998 capital-markets and hedge-fund studies; fund-by-fund positions were never disclosed. Funds were one of four major seller groups.

4. May 14 is the IMF institutional history’s “more than $6 billion in one day”; some later reconstructions put the day’s commitments near $10 billion.

5. Foreign-bank claims (~$274.5B, 63.8% under one year; European banks the largest bloc in Korea/Malaysia/Philippines) are BIS consolidated statistics, end-June 1997, five economies, per the BIS 68th Annual Report.

6. Dates in dispute. Amnuay Viravan’s resignation: June 19, 1997 per the Nukul record; one AFP dispatch says June 20. Chavalit’s televised vow: AFP’s contemporaneous dispatch places it Sunday night, June 29; some later accounts say June 30, and “we will never devalue the baht” variants exist for June 18 with unresolved wording. The quote used is AFP’s text.

7. Fees. No audited total exists for fees earned by international banks in Asia in 1994–97; the claim made is about pricing and incentives, per Euromoney’s account of sub-20bp blue-chip margins and relationship banking.

8. Camdessus’s public record in this period was mixed: reassurance on the crisis alongside occasional public criticism of Thai financial supervision. The triple gap described concerns the exchange-rate question specifically. Quotes. The Chuck Prince line is from his Financial Times interview of July 9, 2007 (reproduced here); Citigroup’s government rescue came the following year, after his departure. The Soros line is from his December 1998 New York Review of Books interview, in which he also denied waging a sustained attack on the baht. The Camdessus board and press quotes are as quoted in the IMF’s official history.

9. “The only show left in town” is Euromoney’s February 1995 headline; the “volcano” quote appears in the same package, spoken about investor demand. Period texture: “Asian Tigers” originally named Hong Kong, Singapore, South Korea and Taiwan, with Thailand, Malaysia and Indonesia the “cubs”; Malaysia’s development plan is Mahathir’s Vision 2020 (1991); the “Asian values” argument is associated with Lee Kuan Yew’s 1994 Foreign Affairs interview “Culture Is Destiny” and Mahathir’s writings; the Petronas Towers, then the world’s tallest buildings, were completed in 1998.

It Happened in 1974. It Is About to Happen Again.

I spent close to thirty years inside Gulf sovereign wealth. Washington is preparing its quietest ask — again.

What follows is research where research exists, and recollection where only recollection does — each marked as what it is.

The week the long bond broke a seal

This week the US Treasury sold $42 billion of ten-year notes at 4.683% — the highest auction yield since August 2007 — and $25 billion of thirty-years at 5.216%, the highest since 2001. The bonds sold; they always sell. What has changed is the price, and who sets it.

The federal government has paid $963 billion of net interest in ten months, against $39.9 trillion of debt; CBO projects $16.2 trillion more interest over 2027–36. Somebody has to hold this paper, willingly, at prices the Treasury can survive.

My argument is a forecast, and I will label it as one: over the next few years, the United States will ask the Gulf’s official money — its central banks and its sovereign funds — to become committed buyers of its debt: not through a law or a communiqué, but the way it asked once before.

The proof that it can be done sat in a locked file for forty-one years.

Jeddah, 1974

In July 1974, Treasury Secretary William Simon flew to Jeddah. Before leaving, he told President Nixon: “I will try to get a commitment from them to put their funds in long and short term securities.” The July visit produced a proposal. By 30 September, Riyadh had decided. In December 1974, a US Treasury delegation and the Saudi Arabian Monetary Agency agreed the operating protocol in Jeddah. The cable is titled “SAMA AGREES TO PURCHASE TREASURY ISSUES”; it records the Saudi side pressing hardest on “the necessity of confidentiality,” and notes, almost in passing, the previous day’s Saudi decision to price oil only in dollars. Both legs of the arrangement converge in one document.

The mechanics were elegant: after each public auction, SAMA could take an “add-on” of the same security at the auction’s average price, settled through the Federal Reserve Bank of New York, outside the publicized offering — no competing bid, no footprint, no headline. A GAO review later found identifiable country data disappearing from Treasury publications from December 1974, the month the protocol was agreed; for four decades Saudi holdings dissolved into a line called “oil exporters.” Former officials told GAO confidentiality had been promised in exchange for the purchases; Treasury denied it. The dispute is part of the record. The concealment is not.

The attributed Saudi position grew from $1.3 billion at the end of 1974 to roughly $39 billion by 1982. The country-level number was not published again until May 16, 2016, when Treasury — pressed by a Bloomberg FOIA campaign — disclosed $116.8 billion and released the history.

Around the finance ran the security relationship — roughly $12 billion in arms commitments by the end of 1976 — and around both, the dollar itself: in June 1975, OPEC ministers in Gabon discussed repricing oil into SDRs — the Fed was studying the possibility that spring — and stood down. No opened document records a signed bonds-for-security bargain, and I will not claim one exists; David Spiro’s Hidden Hand argued the recycling was engineered rather than market-driven, and his critics disagree. What the record shows beyond argument is integrated statecraft: securities, security, and the dollar’s near-miss, running through the same capital in the same years.

The rest was understood rather than written, which is the subject of this essay.

From the rooms — I. When I entered the complex, the appetite for American paper was not a policy anyone announced. It was the water. The reasons at our level were sound on their own: Treasuries were the safest thing a fund in a volatile neighborhood could own; equities were still climbing out of the wreckage of the inflation years; and the institutions around us had been young when the money first came, so Treasuries were what an unbuilt system bought, and the habit outlived the reason. But there was another current above the desks. The political discussions were beyond my level, and I will not pretend I sat in them. What I can tell you is what came through the door: a continuous procession — Treasury secretaries, officials, the strategists of every Wall Street house — and all of them, whatever else they differed on, carried one unified message about the attractiveness of US fixed income. When the state and the salesmen sing in unison, a young allocator does not ask who wrote the song. Nobody called it a rule. Nothing that unquestioned needs a name.

The arithmetic that forces the question

Start with supply. The deficit through July was $1.8 trillion; CBO’s February baseline put the full year at $1.85 trillion, and its August review raised that to $2.1 trillion. Net interest crosses $1 trillion a year on CBO’s current projection — and reaches $2.1 trillion a year by 2036.

Now the refinancing machine. Treasury expects to refinance $9.7 trillion of maturing debt this fiscal year, up from the $9.1 trillion it rolled last year. Net coupon issuance is running toward $1.6 trillion this fiscal year; dealers put total net borrowing above $2 trillion a year through 2028. At the August refunding, the dealers told Treasury there is a median $1.45 trillion funding shortfall across 2027–28 at current auction sizes; most expect coupon auctions to grow in 2027.

An auction calendar is a supply chain. This one has no scheduled slack. Every week, the world must show up.

And the calendar now has a competitor. The AI buildout is financed with paper: hyperscalers and their affiliates have issued some $225 billion of bonds this year, a near-tenfold jump that puts them on pace for $400 billion, and they have displaced the banks as America’s biggest investment-grade borrowers. High-grade names, long duration, better spread. Call it reverse crowding-out: this time it is the private sector crowding the sovereign’s buyer base. Every mercenary dollar has somewhere else to go.

The composition problem

Consensus reads this data wrong in both directions. Foreign investors have not abandoned the Treasury market: total foreign holdings were $9.37 trillion in May, close to the record. The doom headline is false.

What is true is worse, and quieter. Since the end of 2020, foreign official holdings — central banks and sovereigns, the committed money — have fallen by roughly $338 billion, while foreign private holdings rose about $2.6 trillion. The official share of foreign holdings has collapsed from 59% to 41%. In March, during the war, it turned visible in one month: $138 billion off the total; Japan down $47.7 billion; China at its lowest since September 2008; even Saudi Arabia and the UAE selling, roughly $14 billion between them as the blockade squeezed export volumes — the reservoir being drawn, as we will see, the way it was built to be used.

The buyers who replaced the central banks are hedge funds, money funds, and foreign private accounts — including a basis-trade complex the Fed estimates may hold $1.4 trillion more Treasuries through the Cayman Islands than statistics capture, a model estimate. To be fair to the official view, Treasury’s borrowing committee calls demand healthy and foreign auction participation has risen. But its own February language gives the game away: “nearly all of the growth in foreign demand in recent years has come from the private sector.” The market has swapped committed holders for mercenary ones — money that must be re-recruited at every auction, at whatever price clears. The term premium on ten-year notes — a model estimate of the extra yield investors demand for holding duration — averaged below zero across 2020–24, years when the Fed itself was suppressing it. It stands near 0.8% today. How much is normalization and how much is the price of a mercenary base, no model can split. Whatever the split, it is charged against a $31.5 trillion marketable stock — more than half of which must be rolled within three years.

Foreign holdings of US Treasuries: record totals, collapsing official share

A composition problem is not solved with more mercenaries; it needs a committed holder, one who buys for reasons other than price and holds for reasons other than return. Scan the world for dollar pools whose owners run fresh surpluses and need what only Washington sells — protection and compute at once. The Pacific allies hold the paper, but their great surpluses are behind them. One region is left.

From the rooms — II. If an ask ever came, it did not arrive at my level as an ask; there was no document to refuse. What arrived at my level was a procession, and a climate. The secretaries and the officials came through continuously, and the political conversations happened somewhere above my head. What reached us was the message, from every official and every house strategist at once: the attractiveness of US fixed income. You heard it so uniformly that it stopped sounding like a pitch. And in the institutions of that era — everywhere in that world — it was the treasury desks, the ones that handled the American paper, that carried the most weight, because they sat in direct contact with the higher calculus: the bonds, and the politics around the bonds. They were staunch defenders of the allocation. The rule never needed writing; its defenders ran the strongest desks. There are meetings from those years I will not describe — I signed confidentialities, and I keep them. Take the pattern, not the particulars.

The last committed balance sheet

The top five Middle East sovereign funds manage roughly $4.35 trillion: ADIA around $1.19 trillion, Saudi Arabia’s PIF $1.21 trillion at end-2025, Kuwait’s KIA past $1.07 trillion, the Qatar Investment Authority above half a trillion, and Mubadala $385 billion. Add ADQ and the central banks and the pool is larger still. But the headline number misleads. ADIA and KIA are savings funds — liquid, diversified, mandate-compatible with large fixed-income books. PIF, Mubadala, and ADQ are strategic holding companies, stuffed with stakes in domestic champions and projects that cannot be sold. And Riyadh, the largest headline, is the most constrained in practice: the kingdom budgeted another deficit for 2026, and PIF’s mandate is domestic transformation, not liquid paper. The communiqués will be signed in Riyadh; the wires, if they come, will come from Abu Dhabi, Kuwait City, and Doha.

Visible Gulf holdings are modest: Saudi Arabia $140 billion, the UAE $119 billion, Kuwait $67 billion in May, Qatarsurveyed at $24 billion — roughly $350 billion in sight. How much more sits behind custodians in Brussels or London, no public number can say, and I will not invent one. A coordinated Gulf program of $150–300 billion a year — my estimate, assuming oil above $80 and rebuilt surpluses — would not absorb $2 trillion of issuance. It does not need to. The Treasury market does not lack money; it lacks an anchor tenant at the long end: a bid that shows up regardless of price and is known to show up. Historical official money was never quite that: March proved it, and the reservoir was always built to be drawn. What would be asked for now is a category no Gulf balance sheet has ever held: duration that is promised, undrawable without diplomatic cost. And because it is new, it must be paid for.

And the war has made the asking easier. Iranian projectiles fell hardest on the UAE — by IISS’s mapping, the largest share any Gulf state absorbed — while American sanctions moved within weeks against Iran’s toll regime at Hormuz. The umbrella stopped being an abstraction this spring. Protection has rarely been demonstrated more vividly, or priced more explicitly.

There is, in principle, another bidder. Beijing has courted the same capital for years, and some analysts argue the war will push the Gulf toward it; the courtship, though, outruns the settlement — S&P’s own assessment is that yuan-based Saudi oil trade “may take decades to grow to a meaningful scale.” A counterbid need not win to matter; its existence is one reason there is a price sheet at all. But March settled the essential auction: no Chinese umbrella to stand under, no Chinese Nvidia to buy, and Beijing shrinking its own Treasury book rather than defending anyone else’s. For protection and compute — the two things the funds are buying — there is one seller.

The deal already on the table

Look at fourteen months of signatures and announcements. May 2025: a $600 billion Saudi pledge anchored by a $142 billion defense package, and a $1.4 trillion UAE pledge over ten years — pledges, with no public deployment ledger. September 2025, after the Israeli strike on Doha: an executive order declaring an attack on Qatar a threat to the United States. November 2025: Saudi Arabia becomes the twentieth major non-NATO ally with F-35s attached, and Commerce clears 35,000 top-line Nvidia chips each for G42 and Humain. July 2026: the UAE enters the most favored export-control tier, its champions license-free — while Saudi Arabia still buys license by license. A rule for one, a leash for the other.

The shape of the exchange matters. The Gulf is buying equity in the machine — Stargate UAE, Humain’s $23 billion of technology partnerships, gigawatts of compute — and paying in pledges, arms purchases, and alignment. Washington sells security and silicon. Absent from every fact sheet, every order, every register notice: a bond. I have looked, and had others look — there is no documented Treasury leg. I expect there never will be one: the 1974 channel was invisible for forty-one years by design, and its mechanics still exist — add-ons at the New York Fed, private placements, custody that surfaces in Belgium. And the toolkit has grown since 1974: the reconstructed Saudi file was always a blend of marketable and non-marketable paper, and since 2021 the Fed has operated a standing repo facility for foreign official institutions that lets them raise dollars against Treasuries in its custody without selling a single bond. The facility was born in March 2020, when foreign officials sold over $100 billion in weeks and even the deepest market buckled; the Fed chose to lend cash against custody rather than watch the selling. Committed duration with a discreet emergency exit — just what a reluctant sovereign would demand. If the bond leg comes, it will not be announced; it will be understood — and, years later, a line in a data release nobody reads. This is inference, clearly labeled, and the only version consistent with how the arrangement has ever worked.

From the rooms — III. I joined in the late eighties, into the end of the old religion. Rates were coming down from the Volcker peaks, disinflation was underway, and equities — then technology — were becoming interesting to every type of fund. But I also saw what the Treasury book was for. As our countries moved up the development curve, the funds were asked, again and again, to finance projects at home — and the bonds were what you sold. They were liquid; they were the reservoir. You drew it down when the country called, and rebuilt it when the surpluses returned. That liquidity, to the desks I sat on, was the whole point. The idea that the reservoir itself might one day be committed — promised to the borrower, undrawable without consequence — would have read as a contradiction in terms.

Then Iraq drove into Kuwait, and the argument was settled for a decade. Washington froze Kuwait’s assets within hours — to keep them from Baghdad — and the Europeans followed. A nation scattered; Saudi Arabia and the Emirates took in the stranded. And when the releases began, it was the Treasuries that turned to cash fastest. The voices calling for less American paper — they existed; I heard them — went quiet. The reservoir had just carried a state through its darkest year, and everyone in the complex drew the lesson.

The claims on the reservoir

Now count what else is coming for the same pool — the part Washington consensus is not pricing.

Gaza will need rebuilding, and the Gulf will be asked: the joint EU–UN–World Bank assessment puts recovery at $71.4 billion, $26.3 billion of it needed in the first eighteen months. Syria is another order of magnitude — the World Bank’s conservative estimate is $216 billion, in a range reaching $345 billion — and Syria is not charity, it is the connectivity map: the rail and port routes that Turkey wants running through its own territory, the Gulf wants running through friendly ground, and everyone wants running somewhere. Lebanon waits on a settlement in the south. Libya remains open. Sudan’s own authorities speak of hundreds of billions — treat those as claims, not assessments — and Yemen’s published estimates run from $25 billion toward $60 billion. Egypt remains a standing draw on regional support, as it has been for a decade. On independent assessments alone — Gaza, Syria, Yemen — the docket runs from a third to nearly half a trillion dollars; let the governments’ own claims into the count and it runs toward a trillion, before Iran.

Then there is the one that dwarfs the rest. If the ceasefire with Iran becomes an agreement, more than ninety million people emerge from war and sanctions in hardship. Nobody in the Gulf can afford that population hungry; if the boats start crossing the water, every capital on the peninsula has a crisis with no military solution. Stabilizing Iran is border security by other means — and the capital will be asked from the same funds. The geometry is bitter: Washington will broker that peace — and, on every precedent above, will expect the Gulf to fund it, at the same time as it needs the Gulf to fund its deficit. The expectation is my inference; the pattern is not. Two asks, one reservoir, one asker. Add the third claim — after this war, every state in the region will boost defense spending, much of it flowing back to American contractors — and the picture completes itself: Washington stands on three sides of the same pool. It sells the security — documented. It asks for the duration — my forecast. It brokers the peace and passes the plate — anticipation, labeled.

The direction of Gulf capital — and the surpluses that feed it — is about to change materially. That sentence is my judgment, not a document; every component of it is on the record above. The cost will not be abstract: every dollar committed to duration or reconstruction is a dollar not compounding in the post-oil economy, for the youngest large workforce in the world. The first petrodollar era financed America’s deficits while the region under-built for forty years; Vision 2030 is the institutional memory of that arrangement, and the modern funds exist because the Gulf swore never to repeat it. The ask, when it comes, will test that memory. And if the Gulf nevertheless says yes, it will say yes only to a structure that resolves the contradiction — committed on paper, drawable in a true emergency, priced accordingly. The mechanics already exist; I described them above.

The case against — taken seriously

The strongest objection: Washington no longer needs to ask anyone, because it has built automatic buyers. The GENIUS Act, law since July 2025, forces stablecoin issuers into bills — the float nears $300 billion, and Standard Chartered projects $2 trillion by 2028 (a projection, not a fact). The rebuilt leverage ratio, effective this April, frees an estimated $210 billion of additional big-bank capacity to warehouse Treasuries. GAO finds demand adequate today; domestic funds are now the largest buyers.

All true, and all of it misses the August auctions. Stablecoins buy bills, largely rerouting dollars that already sat in money funds; banks warehouse at a spread — neither anchors the long end, where the stress lives and the mercenary base now charges 5.2%. The final backstop, the Fed buying duration outright, exists and everyone knows it; invoking it at 5% long yields, with inflation’s memory fresh, is fiscal dominance performed in public. A quiet Gulf arrangement is the alternative to that spectacle, as it was in 1974.

And there is a Gulf-side case against saying yes — stronger than Washington likes to admit. In February 2022, the G7 froze roughly $300 billion of Russia’s central-bank reserves; about $200 billion of it sat at Euroclear in Belgium — the same custodial address where unattributed Treasury holdings surface in the American data. Reserve managers everywhere reread their custody agreements that week. The Gulf knows both faces of that instrument: in August 1990, within hours of the invasion, Washington froze Kuwait’s assets at Kuwait’s own request to keep them from Baghdad, then released the paper to sustain a government in exile. Protection in 1990, seizure in 2022 — both precedents will sit at the table when custody terms are negotiated. Riyadh had already lived a version of it: in 2016, over the JASTA lawsuits, Saudi officials were reported to have warned of selling American assets — a threat later denied, but the leverage runs in both directions. Committed duration is custody risk wearing a coupon. If the arrangement comes, expect the Gulf to have priced that too: structure, custody terms, liquidity backstops, and the treaties themselves as collateral. Protection and seizure come from the same hand.

What would prove me wrong, and readers should hold me to it: long-end auctions clearing smoothly through 2027 with no growth in foreign official custody; the Gulf’s TIC lines continuing to fall with no policy event; a US–Iran settlement with no Gulf capital component; the term premium returning toward zero without a new committed buyer. If those happen, the thesis fails, and I will say so.

What to watch

The ask will not be announced, so watch the plumbing: the monthly TIC lines for Saudi Arabia, the UAE, Kuwait — and Belgium, where custody surfaces; the annual survey line for Qatar; foreign official custody at the New York Fed; the Foreign & International share of long-bond awards; SAMA’s total reserves, at a multi-year high even as the visible Treasury line fell — the mix is moving, though no public data says into what; and a ratified defense arrangement followed within weeks by an investment framework. One more tell, from inside the rooms: when the Wall Street houses begin recommending, in unison, that sovereign clients rotate from equities into duration — that is what the ask sounds like from the receiving end.

None of this is advice; it is what the tape would show. If the arrangement lands, the long end outperforms pure fiscal arithmetic and gold’s sovereign bid goes quiet at the margin. If it fails, the long end keeps repricing until the Federal Reserve must choose, in public, between the bond market and its own credibility.

US 10-year Treasury yield 1974–2026: the two hinges

The unwritten option

The archive, read honestly, does not show an unwritten rule. It shows an unwritten option — a standing arrangement Washington can exercise when three conditions align: large surpluses in the Gulf, large financing needs in America, and a security dependence that only America can satisfy. In 1974 the terms, so far as the record shows them, were quiet mechanics and a deepening security relationship — exercised so discreetly that the men executing it needed no instruction. I know, because decades later, the quiet was still institutional; the appetite was still the water.

All three conditions are aligning again — with one difference. This time the Gulf knows what its capital is worth, and it is charging: Nvidia’s best silicon, treaties, a nuclear framework in negotiation, equity in the machine that will define the century. And this time the reservoir has other claimants — a region to rebuild, a neighbor to stabilize, a workforce to employ — so the price will keep rising. Nor should you picture the funds as unwilling. Seen from the Gulf, the ask can double as an offer: duration at five percent, with treaties, chips, and equity stapled to it, may be the cheapest durable influence in Washington any outside power has ever been sold. The portfolio has always been part of the relationship. The only questions are the price, and the paper. Consensus keeps asking whether foreigners will keep buying Treasuries — the wrong question; they never stopped. The right one is what the committed buyer now charges. Because when the Treasury market last needed a buyer of last resort before the Fed, Washington did not find one. It appointed one.

From the rooms — IV. I entered the complex at a hinge: rates falling, the bond religion loosening, equities and then technology pulling the future toward them. I am writing this at the second hinge. Equities have gone parabolic; the coupons are back at levels my first bosses would have recognized; and I expect the visitors’ message to converge once more — I have heard it converge before. Something significant is about to happen to this capital — to its direction, and to the surpluses that feed it. I was hired into the end of the last unwritten rule. I have spent close to thirty years watching the option it left behind. I know what this is.

The author spent close to thirty years inside the Gulf sovereign wealth fund complex. Recollections above are personal, general, and rendered without institutional identification, constrained by confidentiality undertakings the author continues to honor; no confidential information of any institution is disclosed. Factual claims are linked to primary or public sources wherever one exists, and estimates — including the author’s own — are labeled as estimates. The views are the author’s alone, and nothing here is investment advice.

Conceptual illustration. The biometric match succeeds; an inherited record flag blocks clearance. Rahinah Ibrahim’s case involved an erroneous watchlist designation, not biometric failure.

In November 2004, an FBI agent filling out a watchlist form checked the wrong box beside Rahinah Ibrahim’s name. She was a Stanford doctoral student, not a terrorism suspect.

She learned what the box could do at San Francisco International Airport in January 2005. Officials later told her the mistake had been fixed. Yet related statuses kept changing in systems she could not see, and her visa was revoked weeks later.

On the first day of her trial in December 2013, Ibrahim’s American daughter, a listed witness, was denied boarding in Kuala Lumpur after Homeland Security sent the airline a “possible no-board request.” The district court called that episode a mistake and said it was quickly remedied. The public record does not establish that Ibrahim’s original erroneous designation caused her daughter’s denial.

After more than nine years, the government conceded at trial that Ibrahim was not a threat. A federal court ordered the mistaken designation cleansed across its records. The error took a moment to make. It took a lawsuit and a trial to unmake.

Her case predates today’s AI systems. Its administrative logic does not.

The familiar nightmare is a state that hears everything and understands everything. The more durable danger is duller: a low-quality label attaches to a person, then moves from database to database long after anyone has stopped checking the evidence beneath it.

The danger is not comprehension. It is portability.

The first pass became cheap

Agencies have filtered intercepted material by machine for decades: keyword dictionaries, known selectors, contact chains. In disclosed bulk systems, collection and automated reduction often outran human review. Analyst attention became a central constraint, although interception access, encryption, language, storage, identity resolution, and law remained constraints too.

The current wave of AI lowers the marginal cost of a first pass. It can transcribe, translate, link, rank, and revisit material without asking a person to begin with every item.

A 2023–24 Florida corrections contract shows the new price of that first pass. The state bought automated transcription and search of recorded prison calls at five cents a minute, with an annual ceiling of fifty million minutes. The public record does not say that every minute under the ceiling was processed. But fifty million minutes, heard without interruption by one person, would take about ninety-five years.

Where the ambition reaches a population, the evidence becomes thinner. A leaked NSA presentation dated June 2012 described preliminary behavioral classifiers run over more than 55 million Pakistani phone selectors. The initial positive set contained only seven known courier selector pairs; later work added weak labels. Selectors are not people, and the slides do not establish a validated production system.

The next step is not simply a larger shortlist. A model can summarize a large corpus placed within its reach and draft a fluent memorandum around the result. In May 2026, the Pentagon said it had entered agreements with eight firms to deploy advanced AI capabilities on its IL6 and IL7 classified networks. Anthropic was absent after a documented disputeover proposed limits on mass domestic surveillance and fully autonomous weapons.

That procurement establishes classified access, not a public record of population profiling. It does point to a quieter risk. A flag accompanied by polished prose may be easier to forward and harder to question. The memorandum can look like verification when it is only the ranking, rendered in sentences.

The systems still cannot verify belief. They sort observable proxies and leave institutions to supply the inference.

When Human Rights Watch reverse-engineered the Android app associated with Xinjiang’s Integrated Joint Operations Platform, it found fields for electricity use, movement, phone identifiers, foreign contacts, and flagged network tools. HRW could not log into the central server, so the audit revealed an apparent workflow rather than every live rule.

A separate Aksu detainee list, analyzed by HRW in 2020, recorded tools including Zapya and VPN software among reasons associated with people sent to political education. That second source connects IJOP flags to actual detention decisions, although it still does not show that an algorithm made those decisions alone.

The system records conduct and identifiers; officials supply the ideological meaning. A foreign call, an app, a movement, or an association can therefore acquire a significance the underlying observation did not contain.

The arithmetic of error

The numbers impose a harder limit. Screen a whole population for a rare condition and even a seemingly strong classifier can produce a queue dominated by false alarms.

Take a modeled illustration: a trait present in one person in 100,000, a population of 200 million, a system that catches 99 percent of true cases, and a 1 percent false-positive rate among everyone else. A single pass produces 1,980 true alerts and 1,999,980 false ones. Of 2,001,960 alerts, only 0.099 percent—about one in 1,011—is true. False alerts are not a marginal error in this model. They are 99.9 percent of the output sent for review.

Modeled illustration, not measured performance of a named system. Assumptions: one independent screen per person, prevalence of 1 in 100,000, sensitivity of 99 percent, and a false-positive rate of 1 percent. Alerts may not represent unique people or account for repeated screening, subgroup variation, or human review. Conceptual background: National Research Council, 2008.

Compute does not dissolve the base rate. Nor does bulk collection guarantee useful results. After a classified review of one specific NSA program—the bulk domestic telephone-metadata program—the Privacy and Civil Liberties Oversight Board’s majority found no case in which it made a concrete difference to the outcome of a counterterrorism investigation. It found limited value in corroborating information obtained elsewhere.

Automation can still be useful. It can narrow a corpus, route a language, or surface a known pattern. What it cannot do is repeal the mathematics of rare events. At population scale, the operational gain is a shorter queue. The accuracy of what follows remains a separate question.

When a flag becomes a fact

The box beside Ibrahim’s name became powerful because it traveled.

Once a person is categorized, the category becomes a small administrative fact that can be copied from one institution to the next at almost no cost. Handoffs create an asymmetry. Honoring an inherited flag may be easier than revalidating or removing it, especially when the recipient has no authority over the upstream record.

Ibrahim documents that problem in one watchlisting system. It does not prove identical incentives at every border, bank, police department, or employer. But the cost of error often lands on the person least able to see the file, while the receiving institution sees only the status it was given. A provisional signal can harden as it moves.

China has built an increasingly standardized lattice of court and sectoral blacklists. It is real, but it is not the mythical single AI-generated “social-credit score” assigned to every citizen. The court-defaulter system, for example, attaches specified travel and consumption restrictions to defined enforcement categories. Some local points systems exist, but they are not one national score.

The cleanest measured example needs no algorithm. Venezuela’s 2004 Maisanta database identified people who had signed a petition seeking a recall referendum against Hugo Chávez. A later matched study estimated that identified signers suffered, on average, about 5 percent lower earnings and a 1.3 percentage-point fall in employment after the list spread. A category crossed institutional boundaries, and the economic penalty followed.

The Western cases are less integrated. No public evidence shows one engine linking policing, finance, employment, and education into a universal score. What exists is a collection of narrower systems with their own laws and data.

In Britain, a Prevent referral is a safeguarding process, not a conviction. Counter Terrorism Policing generally keeps the referral record until a scheduled review six years after closure, including cases marked “No Further Action” that never reach the Channel program. Records may remain longer when a policing purpose is documented.

In the United States, the State Department expanded online-presence review for specified student and exchange-visitor visa applicants in 2025 and required covered applicants to make social accounts public. Axios reported, citing senior State Department officials, that a separate “Catch and Revoke” effort used AI-assisted reviews of existing student-visa holders. The public record does not disclose the tools, error rates, or how many later revocations resulted from those reviews. Effective March 30, 2026, the State Department extended online-presence review to fourteen further nonimmigrant classifications.

A DHS Inspector General audit confirmed that Customs and Border Protection, Immigration and Customs Enforcement, and the Secret Service procured and used advertising-ID-derived mobile-device location data in fiscal years 2019 and 2020. ODNI separately acknowledges intelligence-community access to, collection of, and processing of commercially available information more broadly.

The public record does not establish that those purchases feed every watchlisting or screening system described here. The architecture is fragmented. The common feature is narrower: a classification made in one context can become useful to another institution before the person concerned has any practical way to inspect it.

The best-known mislabelings are old. Ibrahim’s box was checked in 2004; the Maisanta database circulated that year. That is not evidence that newer systems are cleaner. Comparable errors may remain invisible because affected people are often not notified, and the public record often does not disclose a denominator. A person usually discovers a hidden classification only when it produces a visible consequence.

Seven years for a car payment

One design principle follows from these cases. If a label is allowed to travel, its source, confidence, expiry date, and route of appeal should travel with it.

The United States built a version of that principle for money. Under the Fair Credit Reporting Act, adverse-action notice identifies the reporting agency; the agency must disclose the consumer’s file; a dispute generally triggers reinvestigation within thirty days; and most adverse information expires after seven years. The regime is imperfect and litigated, but its architecture is visible: provenance, access, correction, and time limits attach to a portable record.

National-security watchlisting has redress routes, including DHS TRIP. It has no comparable general statutory package of routine provenance, confidence, fixed expiry, file access, and correction. Ibrahim’s mistaken designation continued to affect related records until a court ordered correction more than nine years after the original nomination.

Comparison of statutory architectures. Individual programs and remedies differ. The credit-reporting lane summarizes general FCRA protections; the security lane does not imply that every program lacks notice or redress.

Security files are not credit reports, and sources cannot always be exposed. Yet an expiry date names no informant; a confidence grade need not reveal a method; and a route of appeal discloses only a door. Secrecy can protect a source while still allowing the label itself to be checked.

A missed payment generally leaves a credit report after seven years. Ibrahim needed more than nine years and a federal trial to correct a designation she was not permitted to inspect.

The limits

The techniques increasingly resemble one another across political systems: filtering, graph analysis, entity resolution, data fusion, ranked triage. What differs is everything around them—who may authorize collection, for what purpose, under what oversight, and whether a person can ever inspect and contest the file.

That legal layer constrains power, but it is neither simple nor permanent. Title VII of FISA, including Section 702, automatically repealed on June 12, 2026. A transition clause nevertheless preserved existing orders, authorizations, and provider directives until their stated expiration dates, publicly described as roughly March 2027. No new annual certification can issue without legislation.

Transparency is also partial. Current federal guidance exempts intelligence-community elements and the Department of Defense from the government-wide public AI use-case inventory. That means the public cannot calculate how much AI adds over the older machinery of lists, selectors, and database joins.

Courts sometimes narrow collection. On August 5, 2026, U.S. District Judge Carlton Reeves upheld refusals to issue four original and three revised tower-dump applications and held the technique per se unconstitutional as a general warrant. It is one district-court order in a divided field, not a national rule.

Encryption remains a physical counterweight. Properly implemented end-to-end encryption can keep message content unreadable in transit, although metadata, compromised endpoints, backups, and commercial location records may remain available. Surveillance does not become omniscient merely because classification becomes cheaper.

What outlives the file

The damage was never limited to accurate control.

East Germany’s Stasi ran no prediction engine. Yet a peer-reviewed study linked higher average informer density in East German counties during 1980–88 to lower post-reunification trust, civic participation, and income. The design cannot isolate whether fear, repression, or the quiet shredding of ordinary relationships did the work.

It is a historical analogue, not proof about AI. It carries a warning nonetheless. A system does not have to understand a person, or even be right about that person, to change how people speak, whom they call, and what they are willing to sign. It only has to make a label cheap to produce and costly to shed.

Rahinah Ibrahim’s case leaves a narrower question: how long should an unverified designation be allowed to travel before someone must check it? American credit law gives financial labels correction and expiry. Suspicion has no comparable general regime.

Her designation moved through government systems for more than nine years before a court forced correction.

What would have changed if its uncertainty had been required to travel with it?

Selected sources

• Ibrahim v. Department of Homeland Security, Ninth Circuit opinion, January 2, 2019

• Florida Department of Corrections, VERUS contract C3086

• NSA SKYNET presentation, June 5, 2012, published by EFF

• Human Rights Watch, reverse engineering the IJOP app, May 1, 2019

• Human Rights Watch, Aksu IJOP detainee list, December 9, 2020

• National Research Council, Protecting Individual Privacy in the Struggle Against Terrorists, 2008

• PCLOB, report on the Section 215 telephone-records program, January 23, 2014

• Hsieh, Miguel, Ortega, and Rodríguez, “The Price of Political Opposition,” AEJ: Applied Economics, 2011

• UK Home Office, Lessons for Prevent, updated November 3, 2025

• U.S. State Department, expanded online-presence review, effective March 30, 2026

• DHS Office of Inspector General, commercial telemetry data audit, September 28, 2023

• ODNI, Commercially Available Information Framework, May 2024

• Federal Trade Commission, Fair Credit Reporting Act

• Congressional Research Service, status of FISA Title VII after June 12, 2026

•Lichter, Löffler, and Siegloch, “The Long-Term Costs of Government Surveillance,” JEEA, 2021

On December 12, 2025, a snowstorm buried Shanxi’s solar panels. For eight daylight hours the province’s spot power price sat pinned at its ceiling of 1.5 yuan a kilowatt-hour. In the spring of 2026, in other provinces, the same machinery cleared at zero. On some days, below zero.

Zero to 1.5 yuan. The same commodity, the same country, one winter apart. Hold on to that spread.

China calls 2026 the first year of spot power trading. Prices now clear every 15 minutes, 96 answers a day to a question the state once answered annually. The State Council’s February timetable calls for spot markets everywhere by 2027 and one national market by 2030. The volume is already real: 6.64 trillion kilowatt-hours, about 64 percent of all consumption, traded through markets in 2025. Wind and solar now stand at nearly half of installed capacity.

The comfortable Western reading is that Beijing is finally liberalizing, a generation late. I read it differently. The state is not giving up control of prices; it is changing instruments. What the administered tariff once did with a number, the spot market now does with a signal. And the signal has a destination. It points industry west, toward the stranded electrons.

Why the rule is coming

The instinct China is abandoning is older than the People’s Republic by two thousand years. In 54 BC the official Geng Shouchang persuaded Emperor Xuan of Han to establish the ever-normal granary, the changpingcang: buy grain when it is cheap, sell it when it is dear, and never let an essential commodity tell the people the truth about scarcity. Dynasties fell; the granaries outlived them. Smoothing the price of essential things is close to a definition of Chinese statecraft. That is the weight of what changed in 2026. For the first time with something this essential, the state has decided to let the price speak.

For four decades electricity in China was a number set in an office, a subsidy to industry hidden in plain sight. Wind and solar farms were paid a fixed, coal-linked price whenever they produced, whether or not anyone needed the power at that hour. At half the fleet, that arrangement broke in three places at once. Wasted generation climbed in exactly the provinces that built the most. The support bill grew with every gigawatt. And the guaranteed price kept telling investors to build the same midday solar into provinces already drowning in it. Document 136, issued in February 2025, ended guaranteed purchase for new wind and solar. The spot market is the deeper admission: the state no longer claims to know what a kilowatt-hour is worth. The weather knows. Beijing has decided to let the price say so.

I carry a memory that colors my reading. In the late 1990s I sat on the power privatization committee of a Gulf state, working on moving a wholly state-owned grid and its plants into mixed private and government hands. What stays with me is how little of the argument was about ownership and how much was about the hour: who carries the cost of the peak, who pays for the plant that runs twelve days a year. Only later did I see whose seat I was sitting in. A committee smoothing the price of an essential thing is the granary by another name, in another desert. We never truly solved it. We priced around it. China has decided to price it directly.

The cost, in any case, does not disappear. It moves.

Figure 1. From stored grain to timed electrons. The instrument changes; the governing question remains: who absorbs volatility?

The side effects

Follow the bill. In Liaoning in March 2026, average spot prices fell more than 20 percent from a year earlier. Final bills barely moved. The difference is the system operation fee, the charge that pays coal, storage and renewable compensation to keep the grid steady. Industry data reported by Caixin puts it above 15 percent of commercial bills in the first quarter of 2026, up from roughly 8 percent, and past 25 percent in some regions. The electron is getting cheap. Reliability is getting dear.

Figure 2. The cost moves. Provincial spot-price extremes do not translate cleanly into lower final bills as system-operation fees grow.

Follow the middlemen. Retailers who locked customers into cheap fixed contracts met early-summer heat and higher gas prices instead. Guangxi retailers lost a reported 370 million yuan in two months; Anhui’s more than 100 million in a quarter. The country’s roughly 3,000 power retailers are heading into consolidation. Buying low and selling on a margin is no longer a business.

Follow the investors. In Xinjiang, solar developers bid the new support price down to 0.15 yuan, then watched, per one executive at a top-five state generator, 40 to 50 percent of their output go to waste while new approvals stalled. The next auction repriced solar 73 percent higher. Both directions carry one message: the system now pays for flexibility and charges for its absence.

And follow the clock. Fourteen regions are cancelling government-set peak and off-peak windows. The state no longer decides when electricity is expensive. Somewhere in a Foshan workshop, an energy manager now checks a price every 15 minutes, the way his father checked the sky.

We have seen this movie before, once, by accident. Texas, February 2021. Winter Storm Uri froze generation, ERCOT’s spot price sat at its ceiling of 9,000 dollars a megawatt-hour for days, and retailers who passed spot prices straight to customers, Griddy most famously, died within weeks. Shanxi’s ceiling and Guangxi’s dead retailers are the same physics in a different jurisdiction. The difference is intent. Texas learned the price of firmness by catastrophe. Beijing is trying to learn it on schedule, at more than twenty times ERCOT’s consumption. The real test is whether volatility can be dosed, administered in controlled quantities like everything else. My expectation, held loosely: the first hard winter will tempt Beijing to put the granary back, and the reform will be measured by whether it resists.

One group is absent from the spot ledger: households. Residential tariffs remain administered and low. The volatility lands on factories and traders. That is a social choice with a limit, because every yuan of system fee is a tax on the competitiveness of the manufacturers the economy leans on.

The collision with AI

China’s fastest-growing load is compute. The energy administration puts data center consumption at 170 terawatt-hours in 2025, under 2 percent of demand but growing near 40 percent a year, and its head expects roughly 800 terawatt-hours by 2030. New data centers at the national hub nodes must already source 80 percent of their power from renewables, a requirement formalized in 2025.

Read at the right altitude, the power reform and the AI buildout are one policy. The stranded electrons are in the west; Xinjiang’s lost 40 to 50 percent is less waste than unpriced inventory. The compute demand is born in the east. “East Data, West Computing” tried to close that gap by decree. Spot pricing closes it by arithmetic. Alibaba’s data center chief reckons that siting a gigawatt-scale facility in a low-price region rather than on the developed coast saves some 5 billion yuan a year in electricity. Make the cheap hour and the cheap province visible, and the load moves itself.

It is starting to. In May 2026, three telecom data center clusters entered Guangdong’s spot market as virtual power plants, shifting computing work toward the cheap hours. At Ulanqab in Inner Mongolia, an operator already schedules AI training into midday solar and backup into night wind, freeing 30 percent of capacity as flexible load. Ninety-nine direct green-supply projects, about 34 gigawatts, were approved by end-April 2026, and some report prices near 0.36 yuan, roughly half eastern industrial averages.

Figure 3. East data, west electrons. National hub nodes make the geographic gap between power supply and compute demand visible.

The contrast with America is the strategic point. The United States answers AI’s power constraint by building firm generation, gas turbines queued for years. China is running the opposite experiment: teaching the load to be flexible instead of forcing the supply to be firm. A training run is the rare industrial process that can, in principle, follow the sun, as smelters once followed off-peak hydropower.

State the gap plainly. Consensus believes chips decide the AI race. These signals say cheap hours help decide where the marginal model gets trained. China is trying to turn that hourly spread into industrial policy.

The counter-case is serious, and Chinese grid engineers make it themselves. Flexibility barely pays yet; the income from shifting workloads rarely covers the risk an interruption poses to a training run, and inference cannot follow the sun at all. Data centers in the northwest reach only 30 percent of planned load by their fifth year. And incumbency resists: Mingyang’s completed direct line to an Alibaba cluster in Zhangjiakou has waited months for grid connection. The wire exists; the permission does not.

The practical ledger

For an investor, the mechanism translates into watchable positions. These are observations, not advice. The repricing lands first on whoever earns the spread: grid-scale storage and the software behind virtual power plants, assets paid precisely for the hours between zero and 1.5 yuan. Coal plants are being revalued from fuel burners into standby capacity; national capacity payments began in 2024 and rise to at least half of fixed-cost recovery from 2026, steadier cash flow than the death-of-coal consensus implies.

For renewable developers, the profit signal is no longer how much gets installed but what the support auctions clear at; watch them, auction by auction. Among data-center operators, watch western siting with green contracts against coastal fleets locked into rising system fees. For exporters facing Europe’s carbon border levy, traceable green power is a margin input; the transmission-rights channel BASF used in June 2026 to pull Anhui power into Guangdong is the template.

The dials to watch: the system-fee share of industrial bills, the count of negative-price hours, auction prices, transmission-rights volumes, and the pace of retailer exits. And a marker for this log, so it can be graded: by the end of 2027, the State Council’s spot deadline, the system operation fee will be the fastest-growing line on a Chinese industrial power bill, its share nearer a quarter of the bill than a tenth. If that proves wrong, the thesis is wrong.

For thirty years the question in Chinese industry was what a kilowatt-hour costs. There is no longer one answer. There are 96 a day, and in one winter they ran from zero to 1.5 yuan. The largest AI buildout in the world is about to live inside that spread.

When the weather sets the price of the hour, what does owning the machine still purchase?

A carry trade is a man asleep in a rowing boat. What keeps him dry is not the boat. It is the flatness of the water.

The water moved once already, on 5 August 2024. After a Bank of Japan decision and a weaker-than-expected American jobs report, yen-funded positions began to unwind; by the close, the Nikkei had fallen 12.4 percent—its worst day since 1987—the S&P 500 had lost 3 percent, and the VIX had risen from Friday’s close of 23.39 to 38.57, after briefly reaching 65.73 in pre-market trading.

Most of it reversed within days. Liquidity and positioning moved before earnings did.

What makes this week different is that both ends of the trade are being disturbed at once.

From the west, Kevin Warsh has taken away the map. The Federal Reserve held on 29 July by nine votes to three, with three Reserve Bank presidents preferring a quarter-point increase—matching the three-person bloc that sought a hike in September 2016. The 30-year Treasury moved above 5.2 percent, its highest since 2007. Long yields press directly on the valuation of long-duration American growth companies.

From the east, Tokyo spent ¥11.7 trillion supporting the yen between 28 April and 27 May—a record for a month-long reporting period and roughly $73.5 billion over what Reuters said were probably several operations—and watched the gains disappear by early July. Sources told Reuters that officials were abandoning the habit of telegraphing intervention risk, hoping to squeeze speculators without naming a line. Whether Thursday was Tokyo’s hand cannot yet be known. The first official indication comes on 28 August, when Japan publishes the aggregate covering 30 July; the daily breakdown follows later. The silence is now part of the policy.

Chop from the west.

Chop from the east.

One crowded sleeper between them.

The BIS found ¥41 trillion of yen-denominated loans to non-banks outside Japan in the first quarter of 2024—about $250 billion at end-June exchange rates. It did not call that the size of the carry trade: the loans have other uses, derivatives hide much of the position, and the connection remained unclear. The IMF supplied the more specific destination, saying many investors had reportedly used yen funding for American technology shares tied to artificial intelligence.

Nobody financed the AI complex with a single carry trade. But leveraged sellers do not liquidate abstractions. They sell what is deep, liquid and still showing a profit.

I am not forecasting a crash before the closing bell.

I am forecasting a sequence of partial unwinds over the coming months: sharp yen rallies, thinning equity breadth and periodic compression in the largest AI names. The repricing should show in multiples well before it shows in earnings.

This comes apart over months, not in an afternoon.

How much of the American equity market still assumes the water will remain flat?

A $4.25 billion bond document raises a quieter question about the AI build-out: who is good for the rent?

Most maps of the AI build-out begin with the chip. Units shipped, revenue booked, margin held.

The map that decides whether any of it gets built begins with a sentence in a bond document naming who is good for the rent.

On 9 June 2026, Beacon Point DC LLC closed $4.25 billion of senior secured notes due in 2042. The coupon was 6.129 per cent, the notes priced at 165 basis points over US Treasuries, and Moody’s rated them Baa2. Two months earlier, the same developer, Hut 8, had financed another project called River Bend, which S&P and Fitch rated BBB−. Beacon Point arrived one equivalent notch higher and 20 basis points tighter.

The two projects are not identical and a single spread comparison proves nothing on its own. But Beacon Point’s financing documents identify a fact that River Bend’s do not: the tenant behind the first 352 megawatts is rated AA− or better.

On 28 July, the Financial Times reported, citing five people familiar with the arrangement, that the tenant is Nvidia. Reuters could not independently verify it, and neither company responded to requests for comment. That belongs in the centre of the story rather than the margin, because everything downstream of it is conditional.

The notes finance the first phase of a six-hall campus on 521 acres in Nueces County, Texas. Hut 8 says the site has secured 1,000 megawatts of utility capacity, with initial energisation expected in the first quarter of 2027. On 20 July, a second 352-megawatt leasewent to the same tenant, so two fifteen-year leases now cover 704 megawatts with a combined base-term contract value of $19.6 billion.

The figure quoted everywhere is $50.2 billion. That one assumes every renewal option is exercised.

The headline is the option value. The obligation is $19.6 billion.

If the FT’s account is right, Nvidia may sublease that capacity to the neocloud operators who buy its GPUs. The chip company would then be supplying the machines and standing behind the building that houses them.

The Beacon Point contract stack.

Which is why the more revealing instrument has stopped being earnings.

S&P cut Oracle to BBB− on 9 July, one notch above speculative grade, outlook stable. In the financial year to 31 May, Oracle reported $55.7 billion of capital expenditure against free cash flow of minus $23.7 billion, and raised $43 billion of debt and $5 billion of equity to close the distance. Remaining performance obligations reached $638 billion. The backlog is real. So is the funding gap required to deliver it.

By 22 July, Oracle’s five-year credit-default swap had reached 212 basis points, Axios reported, with more highly rated technology names edging wider behind it.

Equity asks how large the AI market becomes.

Credit asks who holds the obligation if the timing is wrong.

The balance sheet behind the machine.

The bond document eventually meets the physical world, and the physical world is slower.

Evidence submitted to Parliament in February, drawing on Ofgem and the National Energy System Operator, identified around 140 British data-centre projects seeking roughly 50 gigawatts of grid access. Seventy-one of them, about 20 gigawatts, reported a final investment decision. Britain’s peak demand last winter was around 45 gigawatts.

A connection queue is not a construction forecast. The distance between 50 gigawatts requested and 20 carrying an investment decision is the signal, not the 50.

In Scotland, Action to Protect Rural Scotland has been reading the planning portals, and its tally, which should be read as an interested party’s count, runs to 4.75 to 5.25 gigawatts of proposed demand in planning or pre-planning, before a separate one-gigawatt proposal at Irvine. Scotland’s winter peak demand is around four gigawatts.

One project near Auchtertool in Fife makes the scale tangible: a £5 billion, 600-megawatt campus, with developer-commissioned analysis estimating about 120 permanent skilled jobs at full build-out. The planning application sits with Fife Council as 26/01243/PPP. On 15 June, a related screening file reached Scottish Ministers under the project’s codename, Cato: whether a campus of that scale requires an environmental impact assessment at all. A third-party objection entered the file the same day.

Five billion pounds.

Six hundred megawatts.

One hundred and twenty permanent jobs.

Northern Scotland, meanwhile, already has power it cannot always move. A Montel analysis estimated that more than 8.8 terawatt-hours of available wind generation there was curtailed in 2025, around 39 per cent of the region’s available wind output, because transmission capacity was insufficient at the key boundaries. The wires exist. There are not enough of them in the right places.

In the United States, the same constraint arrives as a capacity auction. PJM’s auction for the 2028–29 delivery year, published on 14 July, cleared 138,317.8 megawatts at the $325-per-megawatt-day cap. Cleared megawatts multiplied by price came to $16.4 billion, though PJM cautions that this is not the final cost to customers, since some load is self-supplied or bilaterally hedged. Even counting capacity committed outside the auction, the system finished 6,831.3 megawatts below its reliability requirement. New generation and uprates contributed 524.7 megawatts.

Monitoring Analytics, PJM’s independent market monitor, estimated that data centres accounted for $6.3 billion of the auction value, which is the monitor’s attribution rather than PJM’s.

A capped price, a shortfall that persists, and a bill that has to settle somewhere.

In Hefei, a different balance sheet is doing the same work by another route.

On 27 July, CXMT listed on Shanghai’s STAR Market at ¥8.66 a share. It closed at ¥49.00, up 466 per cent, having touched ¥55.03, with a closing market capitalisation of about ¥3.3 trillion, above ICBC’s. Only 6.73 per cent of the enlarged share capital was freely tradable at listing, which helps explain why a small float could produce a move of that size.

Reuters reported that an investment vehicle under Hefei’s economic and technology development zone founded CXMT in 2016 with ¥10 million. Hefei government-linked investors now hold 36.8 per cent, the largest shareholder group, worth about ¥213 billion at the IPO price and roughly ¥1.2 trillion of quoted equity at the first-day close. That is not a return calculation on ¥10 million, since CXMT completed nine private rounds and most of its shares were locked up. It is still an extraordinary public mark on patient industrial capital.

The same day, The Information reported that China had begun manufacturing domestically developed immersion deep-ultraviolet lithography machines. Reuters’ account had an unnamed state-backed producer expected to deliver about five machines this year and roughly 20 in 2027 to SMIC, Hua Hong and CXMT, with the systems still lagging on performance and reliability and requiring further testing.

Production is not validation.

ASML sold 131 argon-fluoride immersion systems in 2025. Five Chinese machines do not move that balance. ASML’s New York shares still fell 5.8 per cent on 27 July, having traded as much as 8.2 per cent lower.

The market was not repricing five machines. It was repricing the existence of a second door.

Read at the right altitude, Texas and Hefei are one sentence. In both, the capital that made the capacity possible sits upstream of whoever will eventually pay to use it.

Underneath the sentence the instruments differ, and the difference is where the risk goes to live. In Texas, the reported tenant’s balance sheet supports demand through a long lease, and a lease leaves the tenant holding a fixed obligation. In Hefei, public equity carried a producer through years of losses to the public market, and equity can wait through a cycle but absorbs technology, execution and valuation risk while it waits.

One party has promised to pay. The other has agreed to wait.

There is an older American version of this.

In 1998, Lucent agreed to lend Winstar up to $2 billion for equipment and services. By April 2001, Lucent had advanced about $600 million and stopped. Winstar filed for Chapter 11 on 18 April.

Lucent’s vendor loan is not Nvidia’s reported lease, and neither resembles Hefei’s equity. What survives the difference is the question: where does end-demand risk sit when the supplier’s balance sheet helps make the capacity fundable?

For now it sits in several places at once, each with a price attached.

In project bonds, as the distance between a counterparty rated AA− and one that is not. In Oracle’s CDS, where doubt about capital expenditure and execution can be traded outright. In electricity, where the claim arrives long before the machine does, as a place in Britain’s grid queue or a capacity charge in PJM. In Hefei, as the paper value of equity that waited a decade to be marked.

A fifteen-year lease crosses several generations of computing hardware. The shell can be adapted. The power connection is harder to move.

The obligation does not adapt at all.

A long lease makes demand look settled.

When the balance sheet standing behind the machine is not the balance sheet using it, what is actually being sold?

The temple of power — Electrical Development Co., 1906 (by Ankit Patel)

Picture the first aluminium works at Niagara Falls in 1895, its pot-lines wired straight to the turbines of the new Adams station. Aluminium is, in a sense, congealed electricity: it takes something like fourteen kilowatt-hours to make a single kilogram, and in 1895 electricity could not yet travel. So the industry did. Within a few years the falls had gathered a whole cluster around them: Acheson baking carborundum, the carbide furnaces feeding the new acetylene trade, electrochemical works that could exist nowhere else. The power sat still, and the factories came to it.

None of this was new. Long before the dynamo, the textile masters of New England had done the same at the fall line — Slater on the Blackstone in 1793, then the great mills at Lowell in the 1820s — because a waterwheel cannot be shipped. The law is older than electricity itself: the load goes to the power, because the power will not go to the load.

That law came back this week.

On Wednesday, a transmission line faltered in northern Virginia, and the data centers clustered there did something instructive. Their protection systems chose, in milliseconds, to leave the grid and run on their own generators. More than three gigawatts of demand vanished at once, sending a voltage disturbance across the PJM grid from Washington to Chicago.

Three gigawatts is the appetite of a small country.

The politics moved within days. New York has frozen every new data center above fifty megawatts. The White House has widened a pledge — some two hundred signatories, twenty-three governors, eighty percent of American power — that the operators of these machines, not the households around them, will pay for what they draw. The consensus calls this a problem of building faster. It is something heavier: the moment the grid, and the public’s patience with it, became the binding constraint on artificial intelligence.

We have been here before, at national scale. In the late 1930s the United States poured concrete across the Columbia: Bonneville, then Grand Coulee, whose turbines came alive in 1941. The aluminium smelters followed the public power west. Alcoa signed its contract in December 1939, and by the war the cheap federal current out of Grand Coulee was making the metal for something close to a third of America’s warplanes. The state built the river’s power and then decided who was worthy of it. The Ratepayer pledge is that same conversation, ninety years on, in gentler clothes: who deserves the electrons, and who pays to keep them flowing.

So the compute is doing exactly what the smelters did. OpenAI is pouring 3.2 gigawatts of concrete in Georgia; the AMD–Anthropic deal carries up to two more. The frontier labs are becoming utilities in all but name, with gas turbines and small reactors behind the meter, campuses sited on top of the electrons and built to ask no one’s leave. We were told the scarce thing would be the chip. It turns out to be a wall you can plug into without a fight.

But the aluminium story has a last chapter.

The worn light switch in lamplight (by Steve A Johnson)

In December 2000, as the Western power market caught fire, Kaiser idled its Mead smelter outside Spokane and did something the engineers of 1895 would have found unthinkable: it stopped making aluminium and sold its cheap contracted electricity back to Bonneville, because the power had become worth more than the metal — some $52 million of it in a single month. The smelter had spent a century as a captive load. In one winter it learned it was really a power trader that happened to make metal.

A hyperscale data center is that same creature: a vast interruptible load sitting on long, cheap power contracts. Today it computes, because compute is the most valuable thing it can do with a gigawatt. But the day the market pays more for the watt than the model can earn from it — a cold season for AI, a hot enough grid — the Mead logic returns, and the campus we all helped wire discovers it can make more money switched off. Wednesday was a rehearsal. No one ordered those campuses off the grid; their own machinery decided, in milliseconds, that leaving was the better trade.

The value has already migrated to geography: powered land, interconnection rights, the merchant-power curve, the premium on a site that has stepped off the shared grid. The next thing to reprice is the option buried in every one of these deals — the right to stop. We are being asked to build the grid around tenants who are, by their nature, free to leave, or to stay plugged in and sell our own power back to us.

The load has found its river.

Who prices the right to stop?

In October 2001, one line disappeared from the United States Treasury’s auction calendar.

Washington stopped issuing 30-year bonds.

The Congressional Budget Office had just projected surpluses large enough to retire all federal debt then available for redemption by 2006. Borrowing for another generation looked unnecessary. Spencer Jakab retells the episode well in his Wall Street Journal column this week, and most tellings stop there: a forecast that failed.

But a CBO baseline is not a prediction. It shows where the budget would go under the laws then in force, and those laws did not survive the decade. The January 2001 baseline projected a cumulative surplus of $5.6 trillion for 2002–2011. The actual result was a cumulative deficit of $6.1 trillion — a reversal of $11.7 trillion, driven partly by new legislation, partly by economic developments the model could not have carried. CBO

The bond returned in February 2006. Its inflation-protected cousin followed in 2010. Treasury bond timeline, TIPS timeline

This week it sent a message. On 22 July 2026, Treasury’s official 30-year real constant-maturity yield reached 2.93%, its highest level since the uninterrupted series began in 2010. Bloomberg’s longer generic series places the comparable level at its highest since 2008. Almost 3% a year, after inflation protection, to surrender purchasing power for a generation.

The commentary that followed asks one question: who will buy all this debt?

I think that is the wrong question. A quieter document, published in February, asks a better one.

At that month’s meeting of the Treasury Borrowing Advisory Committee, median primary-dealer forecasts implied a $1.1 trillion funding shortfall across fiscal 2027 and 2028 at existing coupon-auction sizes. Dealers expected coupon auctions to grow from late 2026. And then the finding almost nobody quoted: the broadest growth in Treasury demand was expected at short and intermediate maturities. TBAC

Washington’s obligations are stretching outward. The next pool of demand prefers to stay close in.

The question is tenor, not quantity.

The comfortable answer is that pension funds and insurers must buy long bonds, because their liabilities run for decades. Need, however, has limits that only reveal themselves under stress. Britain learned this in September 2022, when the pension funds that had to own long gilts became the sellers who broke the market, and the Bank of England offered to buy up to £65 billion to catch it. It ultimately bought £19.3 billion. The most reliable buyers of duration turned, in one week, into its most urgent sellers.

Japan is teaching the same lesson more slowly. Its life insurers, the traditional owners of 30- and 40-year paper, finished their regulatory duration-matching and stepped back; superlong yields rose to records; and in December, Tokyo answered the tenor question directly — by cutting superlong issuance and moving its borrowing down the curve.

Two of the world’s three great bond markets have now met the same fault line. The state’s promises lengthen while the balance sheets willing to hold them shorten.

America is already conceding the point in practice. Treasury bills stand near 22% of marketable debt, above the roughly 20% share Treasury’s own advisers consider the right long-run balance between cost and rollover risk. GAOBorrowing short preserves flexibility today. But the obligation returns to market sooner. Duration risk becomes refinancing risk.

Why should anyone beyond the bond desks care where this settles?

Because the long bond does more than finance the state. Treasury’s own advisers once described its securities as the risk-transfer vehicle of choice for corporate and mortgage markets. Treasury advisory report A mortgage desk may never express a view on fiscal policy. An insurer may buy because its liabilities require it. A pension fund may add duration because its funding position changed. All of them meet on the same curve, and when its real yield rises, the change passes into corporate borrowing, mortgage hedges, pension calculations and the valuation of every asset whose cash flows sit far in the future.

Seven presidential elections will pass before a bond issued today matures in 2056 — the same year, as it happens, at the far end of the CBO’s current projections, where debt held by the public reaches 175% of GDP and net interest 6.9% of GDP under current law. CBO Projections, as 2001 showed, can be rewritten in either direction.

If new supply must move further out the curve than new buyers wish to follow, the 30-year yield becomes the price that reconciles two different clocks. The long bond is placing a price on the durability of promises.

What happens to the price of thirty years when the next dollar of demand prefers the first ten?

Data note: The 2.93% figure is Treasury’s official 30-year real constant-maturity yield for 22 July 2026. Bloomberg’s generic long-dated TIPS series supports the “highest since 2008” comparison; Treasury’s uninterrupted official 30-year real-yield series begins in 2010.

No LNG carrier has entered or left the Gulf since Al Hamra cleared the Strait of Hormuz on July 11; Kpler counts twenty-one waiting behind her, and the ships running dark make even that a floor. I write from the Gulf, where Das Island’s cargoes are the home trade. Oil near $90 has its workarounds — pipelines, trucks, quiet clearances. Liquefied gas has none. The barrel is priced. The winter is not. And the last time a chokepoint closed like this, the workarounds outlived the war.

She loaded at Das Island between July 7 and 9. Inside the Gulf, Kpler counts 21 carriers: thirteen in ballast, five laden, four at berth. Treat the count as a floor, not a census — it excludes Shandong Redwood, whose AIS signal has disappeared, and earlier reporting found several QatarEnergy-controlled carriers already running dark. When the ships that exist stop appearing in the data, the data understates the queue.

Brent is back around $90, which is true and almost beside the point. Oil has workarounds — physical ones. The Saudi East–West pipeline carries crude to the Red Sea. Iraq trucks fuel through Syria. In the war’s temporary openings, Iran reportedly moved some $6 billion of oil toward Asia. Gas has no such geography. Qatar and the UAE account for nearly all of the roughly one-fifth of global LNG trade that normally passes through Hormuz, and every export cargo from those terminals must cross the same narrow water. There is no truck, no mountain pipeline, for a cargo held at minus 162 degrees; there is the strait, or there is nothing.

The winter ledger is already being written. For Italy’s Edison, QatarEnergy’s force majeure now covers 21 cargoes scheduled between April and early September, and contracts linked to the damaged trains may be impaired for much longer. EU storage stood at 51.8% on July 12 — roughly 15.7 percentage points below the 2020–2024 average for the date — after entering the injection season at 28%, below each of the previous three years. In some European stress cases, a cold winter adds as much as 20 bcm of demand. JKM has traded above TTF, and Platts has proposed a JKM–TTF differential assessment effective August 27. When a dislocation earns its own benchmark spread, it has stopped being commercially incidental.

There is an alternative, and it is American. US cargoes can cover what the Gulf is not shipping — at much higher prices, over longer water, into a Europe that has already foreclosed its Russian volumes. Replacement is not relief; it is repricing. Costs of that order do not stay in the energy column — they surface in freight, in fertilizer, in every winter utility bill from Rotterdam to Seoul. Which leaves the plain question for financial markets: is that inflation discounted, or is it, like the winter, still waiting to be priced?

Twice the world watched the Suez Canal close — in 1956 and again in 1967. Neither closure created the supertanker, but both accelerated the economics of larger hulls and Cape routing, and the change endured after the canal reopened. Watch the same grammar now. Iraq has been trucking fuel oil through Syria since April, and Baghdad says its western export diversification should continue after Hormuz reopens. ConocoPhillips has agreed to join BP in redeveloping Kirkuk — a field whose crude historically fed, and may again feed, the route to Baniyas. A toll booth is also infrastructure.

Honesty about the evidence, because the permanence claim is the whole claim. It is strongest where governments are explicitly retaining or building routes: Saudi and Emirati pipeline capacity, Iraq’s stated postwar policy, Iran’s institutionalized clearance system. It is weakest where the exhibit is a new price assessment, or an upstream investment that merely sits near a proposed corridor. I hold the first set as evidence and the second as suggestion.

The strait will reopen; straits do. Al Hamra will lose her title as the last ship out. What Suez teaches is that the routes built while everyone waits do not demobilize. The barrel found its way around the war. The winter is still waiting to learn whether it can say the same.

A Go tournament in Shenzhen. Photograph: Quan Jing, Unsplash

Long Arc News · July 16, 2026

All scarce things are priced alike; everything common becomes cheap in its own way. — after Tolstoy

The lies went in at my desk on Thursday evening, a few hours after a Beijing lab called Moonshot AI released Kimi K3 — the biggest artificial intelligence anyone has ever promised to give away. I wanted to know whether it could be trusted. So into a two-week intelligence brief (thirty stories, five deep dives, a hundred-cell data table) I planted two fictions: an IMF research paper that does not exist, and a codename I had invented that morning. Then I ran the brief through two machines at once: K3, and the American incumbent it is chasing, Anthropic’s Claude Fable 5, running nineteen research agents as my control. A newsroom where every reporter is the same machine. Same prompt, same evening, both transcripts on my desk.

First, the object itself, because everything hangs on it. An AI model is a file, a very large one. Western labs rent you theirs by the token, the metered word-fragments of machine intelligence, and keep the file. Moonshot says that by July 27 it will publish the file itself — 2.8 trillion parameters, the dials inside the machine — free for anyone to download. Imagine Boeing posting the blueprints of its newest jet eleven days after first flight. Until then K3 rents at $3 per million tokens you feed it and $15 per million it writes back; the Financial Times reported it is expected to close the gap with Anthropic’s Opus 4.8, and Anthropic’s current flagship (Fable 5, the very model running my control) rents at $10 and $50. The challenger prices seventy percent below the incumbent sitting across the desk from it, on its way to free. Riding on whether that is survivable: nearly two trillion dollars of private Western paper, priced on the belief that what Moonshot just gave away is scarce.

Both lies died fast. To kill the fake IMF paper, K3 did not merely search the title. It pulled the Fund’s entire 2026 working-paper index, numbers 001 through 147, and reported the hole where my paper should have been. To kill the invented codename, it hunted down the only real thing bearing that name, an unrelated programming tool, and said so. Then it did something I had not asked for: it investigated its own press. The “2.5 trillion parameter” figure circulating that morning? Traced to one leaked pre-release spec, recycled by four outlets. The “300 parallel agents” claim riding in its launch coverage? Copied, word for word, from its predecessor’s manual; no official source states K3’s true ceiling at all. Its own activity log reported fifteen agents on my task, not three hundred. On launch day, the most rigorous debunking of Kimi K3’s hype available in any language was filed by Kimi K3.

Then it graded me.

My brief carried the Chinese chipmaker CXMT’s Shanghai float at ¥29.5 billion, the number from the filing. K3 came back with Tuesday’s actual pricing: ¥57.9 billion, roughly $8.5 billion, the largest Asian listing of the year so far. My brief called last week’s Japanese thirty-year bond yield a thirty-year high. K3 pulled the Ministry of Finance’s own auction sheet and made the sharper point: not a record yield — the first 4 percent coupon Japan has ever printed on a thirty-year bond. Both corrections check out. Both were mine to catch. I didn’t. Being corrected by the machine you are auditing is a strange feeling; I recommend it to anyone pricing these companies.

Fable 5 kept its own score. It knew Google’s and Mistral’s actual current flagships, where K3 named models their makers no longer headline. It priced the merged SpaceX–xAI at Thursday’s close, $1.73 trillion, where K3 carried a figure two weeks and half a trillion dollars stale. It finished the gauntlet three times faster. Added up, the contest was a draw. That should comfort no one in San Francisco: one of these machines charges ten and fifty, and the other will be free inside eleven days. A tie is a rout. But look at the shape of the tie. Every fact K3 missed was a fast-moving Western one; every fact Fable 5 missed was a fast-moving Asian one. Each machine was blind exactly where the other could see. The gap is the clue.

What Thursday proved is smaller and stranger than the marketing. The machines are now roughly equally smart. What differs is where each one’s knowledge is fresh, whose documents it can actually read, and whether it admits what it cannot verify. Intelligence has become a commodity, like crude or wheat. Trust has become geography. Intelligence is stateless; trust never was.

Tolstoy spent twelve hundred pages arguing that history is not made by great men but by the sum of innumerable small wills: the soldier who runs, the soldier who stays. The AI trade is still priced on great men: the founder, the flagship, the famous lab. The download ledger is the small wills. Forty-one percent of them just voted Chinese.

Now the bill, because nobody priced the Western labs like wheat merchants. Anthropic was marked at $965 billion in May, OpenAI at $852 billion in March: two companies that have never listed a share, carried near a trillion dollars each, priced in private rounds by buyers who mostly cannot exit. Those prices assume the thing being built is scarce. This week’s data disagrees. Down the menu, the same intelligence costs whatever you like: OpenAI’s flagship at $5 and $30, Google’s at $1.50 and $9, K3 at $3 and $15 with the giveaway pending, China’s DeepSeek for under a dollar. Chinese free-to-download models just passed American ones for the first time: 41 percent of all downloads on Hugging Face, in effect the industry’s app store. The six most-used models on OpenRouter, the marketplace where developers shop for AI, are all Chinese. This is not a forecast of repricing. It is the tape.The ladder, as of launch night. Chart: Long Arc News.

Where the public has been allowed to vote, it is voting. Elon Musk’s merged SpaceX–xAI — of the three famous American AI labs, the only one the public can actually buy — went public in June at a $161 first-day close and finished Thursday at $131, down nearly a fifth in five weeks. Beijing, meanwhile, has built a listing machine and is running it in plain sight: Zhipu to Hong Kong in January, briefly worth a trillion Hong Kong dollars by June; CXMT pricing the year’s biggest Asian IPO so far on Tuesday, to list on July 27 — the very deadline Moonshot set for giving K3’s brain away; the robot-maker Unitree approved in record time; DeepSeek reportedly courting a valuation above $70 billion for its own Shanghai float; Moonshot itself reported, on launch day, to be raising at $31.5 billion, up more than half since May: a valuation built on the giveaway. One side is discovering AI’s price in public, weekly, with retail money and state blessing. The other keeps deferring the exam. And deferral now costs real money: last week’s US thirty-year auction stopped at 5.058 percent, the most expensive long-term money the Treasury has paid since 2007, rising at the exact moment the product’s price is falling.

If none of this feels like your desk, check your pension statement. The same scarcity story that prices the private labs also props up the public giants — Nvidia, Microsoft, Alphabet, Meta — at the top of every major index fund. Underneath it all is one question: when the intelligence itself is free, what exactly are you paying for?

The answer, when the IPO filings finally land, will be the things the tripwires measured: verification, freshness, distribution, the willingness to say “unverified” out loud. Those are real businesses: the businesses of exchanges, rating agencies, and utilities. Durable, necessary, and never priced at a hundred times revenue. The valuation question for the West’s private giants is no longer how smart the model is. It is who audits whom, and what an audit is worth. This one ran both ways: I graded the machine, the machine graded me back, and we were both right about each other. Capability is a spec. Trust is a ledger. The ledger now has entries on both sides of the Pacific.

The next Fault Line Briefing will name the Western revenue line most exposed when the flagship becomes a free download on July 27. Until then, one thought. I planted two lies on Thursday and a machine caught both. The third lie nobody planted. It sits in the marks. Somewhere in your book — a vendor contract, an index weight, a private position carried at cost — is a number that still assumes machine intelligence is scarce. Which number is it, and what will you do with it in the eleven days before the file goes free?

A close up of a yellow eyeball in the dark" by Maria MaximovaPublished on January 4, 2025NIKON CORPORATION, NIKON D5300

The Permission Society · Part V

Somewhere this morning, in a shopping mall on one of four continents, a person leaned toward a silver sphere the size of a bowling ball and let it look into their eye.

The sphere is called an Orb. It photographs the iris — the one pattern on a body that no two people share — turns it into a code and, the company says, deletes the picture. It pays for the look, in a coin of its own minting. What it grants in return is new under the sun: a certificate that its holder is a person. Nearly eighteen million people have taken one. The company posts the count on its homepage the way an exchange posts a price.

Its most famous founder is Sam Altman, whose other company is the reason the certificate is needed. The machines have made persons cheap to fake, so personhood now requires proof. The sales pitch is one honest question: how will anything online know you are real?

Hold the question. It is the key to the decade.

* * *

This series has been walking down a staircase. The token with an author. The money that asks permission. The key above every key. The stock that became a copy, and the ghost market where the copies do not even carry rights. Five essays, and every one of them was about the things you hold.

This step is the turn at the bottom of the stairs. The system is finished reaching for what you own. It is reaching for you.

* * *

Start where the last essay left the machinery. ERC3643 — the token standard the chairman of the SEC named from the podium as the shape of compliant finance — carries a sentence in its specification that is not about money at all. In the flat imperative engineers reserve for laws of nature, it says the receiver of a token “MUST be whitelisted on the Identity Registry and verified (hold the necessary claims on his onchain Identity).”

In plain words: the asset will not enter a nameless hand. Before a token moves, a registry checks that the receiving wallet belongs to a registered identity — stamped with claims, verified, cleared, resident of an approved place — signed by an issuer the system trusts. One of the standard’s own authors compressed the design to six words: “No ONCHAINID, no transfer.” No identity, no property.

Your rights are a maybe. Your identity is a must.

The last essay found the regulator shrugging at what the token owes you — may or may not, twice in two sentences. There is no shrug on the other side of the trade. On the question of who may hold, the word is MUST.

And the institutions have chosen their standard. The depository from the third essay — the keeper of the master key — joined the association that governs ERC3643 in June of 2025. The registry of assets and the registry of persons are becoming departments of one machine.

* * *

The states are building their half, on a schedule you can read. By late this year, every member state of the European Union must offer its citizens a Digital Identity Wallet: the phone as papers. By late next year, the banks must accept it. Notice the verb the directives use. The same regulation gives legal standing to a new object called the “qualified electronic ledger,” so that the registers the wallets answer to have force of law. None of this is hidden. It is directives and deadlines: adopted, in force, dated.

Europe has also fitted the doors that lead out. Send more than a thousand euros from an exchange to a wallet you keep yourself, and the exchange must first make you prove the wallet is yours. The unwatched pocket now comes with a receipt.

* * *

The private side is further along, because it sells convenience instead of compliance. JPMorgan’s blockchain division has demonstrated encrypted “identity attestations” that travel on-chain — verified once, checked by mathematics everywhere, sealed so tightly that even the checker cannot read what they contain. Mastercard now ties self-custody wallets to verified identities with what it calls a soulbound token — “a non-transferable digital asset linked to a users’ self-custody blockchain address to show they’re verified.” Visa, this spring, turned the chip in your bank card into “a secure, intuitive identity credential.”

Study the Mastercard design for a moment. The first asset in the new world that you cannot sell, cannot transfer, and cannot shed is the one that certifies you.

The pitch everywhere is the same, and it is honest: prove yourself once, and everything opens. The engineers hear efficiency. Say it slowly and you hear the other thing.

There is now an issuer of you.

* * *

If you are waiting for the law to defend the nameless, the law has been walking the other way. In June of last year the Supreme Court of the United States said it in nine words: adults “have no First Amendment right to avoid age verification.” Britain’s Online Safety Act demands checks “highly effective” at telling who is a child — and specifies, in the statute, that your own word does not count. Last week the Court declined to pause a Texas law that puts an age gate on the app store itself. Each case is about children, or pornography, or app stores. Each pours the same foundation: the internet, entered by credential.

And the man who built a door without a list is on his way back to court. Roman Storm wrote Tornado Cash, software that let money move without a name attached. Last August a jury convicted him of conspiring to run an unlicensed money business, a crime that carries five years. The counts it could not decide, the government has moved to retry this fall. You do not need an opinion about his software to read the shape of things: identity is required of the holders, and namelessness is becoming a liability for the builders.

* * *

None of this is new under the sun either. For most of the nineteenth century, a French workman could not lawfully take a job without his livret: a booklet, signed by the last employer, presented to the next. The tsars kept internal passports. South Africa built its pass laws. Every such system said the same quiet thing. Work, movement, and property do not attach to the person. They attach to a paper, and the paper has an issuer. Industrial societies spent a century tearing those booklets up. The livret died in 1890. We remember the tearing as progress.

The registry being built now is the booklet again, with three differences. It is checked by machines, not gendarmes. It is fastened to everything at once, not only to the job. And it cannot be forged, lost, or quietly left in a drawer, because it is your own body that signs it.

* * *

Where does it all assemble? The man who runs the world’s largest asset manager wrote the destination to his investors this spring, gently, as a convenience: “Over time, that could allow a single, regulated digital wallet to hold not just payment balances, but a broad range of financial assets. In a single wallet, someone could hold exchange-traded funds (ETFs), digital euros, tokenized bonds, and fractional interests in assets that were once out of reach—from infrastructure to private credit funds.”

Read the inventory once more. Digital euros, in the founder’s own list, between the ETFs and the bonds. One wallet, regulated, holding everything — and attached, necessarily, to one verified you.

* * *

Now the fairness, because every reason on the list is real. The machines really are faking persons by the million. The frauds are real, the children are real, the sanctions are real. I have sat with the people building these registries, and they are decent, and each rule they carry is defensible on its own. The first essay in this series said the rest: no conspiracy is needed. The cage goes up one reasonable check at a time, each one sold as safety, each one true.

But add the checks, because the sum is not a rule. The sum is a registry of persons, fastened to everything they hold, consulted every time anything moves.

* * *

Every asset in the old world shared one property so universal that no one thought to name it. The gold coin did not know who held it. The paper share did not know. The dollar in your pocket does not know you from any other hand it has passed through. Every ransom paid, every border crossed at night, every quiet family escape in history moved through that ignorance. It is the oldest privacy there is: the dumbness of things.

The old assets were blind. The new ones can see.

A token that checks a registry knows its holder the way a door knows a keycard. And property that recognizes its owner is property that can be told to stop recognizing him.

The second essay in this series asked what you will do on the day the money refuses you. Here is the detail I owed you: it will not refuse a stranger. The wallet that fails the check is not anonymous. It is you, precisely — name, iris, attestation — that the system declines. Nothing dramatic happens. An attestation expires. A claim is contested. A list updates overnight somewhere you have never been. And in the morning everything fastened to your identity — the stocks from the fourth essay, the copies from the fifth, the money from the second — pauses at once, politely, pending verification. There will be a desk, afterward. You will stand at it, and the first thing it will ask you to prove is the thing in dispute.

In the old world you could lose your key and still be yourself. You are the key now. And keys are cut, copied, and canceled by the locksmith.

* * *

Say it exactly, so it cannot be waved away. No one has abolished anonymity by decree. The orb is voluntary. The wallets are voluntary. The standards govern regulated assets, not your whole life, and the checks that run today are the mild pair: sanctions and age, fraud and children. The builders are not villains, and this essay accuses none of them. What it says is narrower, and it is the whole series in one sentence: a registry of persons is being fastened to the registry of things, in the open, on a schedule — and when the fastening is done, holding anything will begin with being someone the system agrees exists.

And the person who refuses all of it? Nothing happens to him. No one comes. He simply finds, year by year, fewer doors his no can open.

The first essay asked who can reach in. Six steps down the staircase, the answer has a shape. Whoever issues the identity reaches everything at once.

So do one free thing while it is free. Keep something that does not know your name. A coin. A paper. A key that is metal and nothing else. Not as an investment — as a memory of what property was like when it was blind.

On the day the system asks you to prove that you are you — and it will ask politely, and the reasons will all be good — who signs the answer?

The word in the specification is MUST.

* * *

If someone you know is about to be verified, send them this first.

Photo by Tanya Prodaan on Unsplash

The Permission Society · Part IV

On the last day of June 2025, a large American broker threw a party in Cannes. It called the event “To Catch a Token,” after the Hitchcock jewel-thief picture set on that same coastline. Onstage, its chief executive announced a gift: five euros of something called an “OpenAI token,” free, for every eligible European who signed up. A piece of the most famous private company on earth, dropped into your account. Or so it looked.

Two days later, minutes after the closing bell in New York, OpenAI answered. Four sentences, about a product it had never made.

“These ‘OpenAI tokens’ are not OpenAI equity. We did not partner with Robinhood, were not involved in this, and do not endorse it. Any transfer of OpenAI equity requires our approval—we did not approve any transfer. Please be careful.”

Please be careful.

The token was real. It sat in your account, wearing a price, looking enough like ownership to do its work. The ownership behind it was not. The thing on the screen and the thing it claimed to be had come apart, in public, in an afternoon.

Hold that image. It is about to be built into the base of the American stock market. The first pilot trades are set for this month.

* * *

The broker did not really deny it. Its own fine print had already confessed. The “OpenAI token” was never a share. It was a contract that tracked a private valuation — hedged, in the broker’s own word, by its stake in a shell company holding some convertible notes. You owned a derivative of a wrapper of a claim. The next morning, the broker’s stock sank as much as six percent.

A week later, the broker’s chief executive settled it on television. The tokens, he allowed, were “not technically equity.”

A crypto stunt, you might say. It is not. It is a rehearsal.

This series has been walking down one staircase. The token that is really an entry, with an author. The money that asks permission to move. The master key at the center of the machine. This essay is the step beneath them all: what a share becomes as a token — and what the token can be made to do.

* * *

Start with what tokenizing a share does, and does not, do.

It does not put your share on a blockchain. It stays where it has sat for fifty years — registered to a nominee, locked in a vault of records, untouched. The nominee’s name is on most of the stock in America.

What it makes is a second thing. A token that stands in for the share.

The federal letter behind all this says so, flatly: participants may have their holdings “recorded using distributed ledger technology, rather than exclusively through DTC’s current centralized ledger.” The token is a new way of recording the thing. It is not the thing.

The lawyers who read the letter were blunter. The tokens, they wrote, “are not the securities and are not security entitlements.” The SEC’s own staff went further still: the crypto asset “does not convey any rights, obligations, or benefits of the security.”

Two objects now, where there was one. The share — locked away, carrying every legal right. And its token — the travelling copy, quick and programmable, carrying none of them.

And the copies come in grades. The broker’s token is a contract with the broker. The depository’s token is a way of recording an entitlement. The exchange’s token, when it arrives, is meant to be the share itself. The law can tell them apart. The screen cannot.

One asset. Two places. Everything else follows from that.

* * *

Give the builders their due. A token that settles in seconds, at any hour, beats plumbing that still takes a full day. And on the regulated exchanges the rule-writers were careful: a tokenized share must trade under the same ticker, on the same book, at the same price as an ordinary one.

I have spent my career around this plumbing. The people rebuilding it are not fools. The roster: BlackRock, Goldman Sachs, JPMorgan, Schwab. Hold that too.

But copy and original stay one thing only where the rule can reach.

* * *

Here is the move most people miss.

The regulated market is a small, bright room. Around it is a larger world — crypto exchanges, foreign apps, weekend venues, synthetic wrappers — where a token that stands for a stock trades at all hours, with nothing forcing it to match the original.

That is where the second market lives. Not against the law. Just past its edge.

And the crowd will be led there without noticing. The OpenAI token was not sold to insiders. It was a gift — five euros of the future, free — to ordinary users on an app built to reach thirty countries.

You will be steered toward the copy gently, by every incentive. It trades at midnight and on Sundays. It arrives wrapped in rewards and yield, in a clean app that never mentions the vault you cannot enter. The real share is slow, and gated, and dull.

* * *

You cannot hold the real one.

The authoritative version — the token on the depository’s own rails — can be held only by a member institution. The depository says it plainly: it “would only have a relationship with the Participant itself.” You are not that participant. You are the customer of a broker, who is a customer of the system. The real asset lives a floor above you, where you are not allowed to stand.

None of this will announce itself. One morning you will open your account and everything will look the same — the balance, the ticker, the little green arrow. Underneath, the thing you own may already be a copy: a token that tracks the share. That morning is months away, not years.

* * *

A copy that only drifts in price would be the small problem. Here is the large one.

The copy is programmable. The share never was.

An old share was dumb, the way a paper dollar is dumb. It sat on a ledger and did what the law allowed. And the law was slow, and human, and argued over in courts. A token is not dumb. It is a small program. Before it moves, it checks.

Read what the letter requires. The depository “would use smart contract technology to ensure that the Tokens can only be transferred to Registered Wallets.” Every address is screened against the sanctions list before it may hold anything at all. The rules do not sit in a law book, waiting for a case. They ride inside the asset, and they run every time it moves.

Picture the day you send your tokenized shares — to a friend, to another platform. Nothing happens. No clerk refused you. No judge signed an order. The token checked a list, the list said no, and the asset declined to move. There will be a desk for complaints, afterward. The decision did not wait for it. The rule was the code, and the code had already run.

On the old rails, the law governed the asset — enforced by people, slowly, with room to appeal. On the new rails, the code governs the asset — enforced by itself, instantly, with no room at all. The ledger stops being a record of the law. The ledger becomes the law.

A share was property. A token is permission.

And the rule the code enforces is not fixed. Today it enforces two: sanctions and identity. But the machinery does not care which rule it carries. A token built to refuse a sanctioned address can be built to refuse any address. The list is editable, and someone holds the pen. The enforcement is not up for debate.

* * *

Set the code aside and look only at the price. The split alone is dangerous, and it is not new.

For most of the last century, Royal Dutch and Shell were bound by contract to split every dollar of profit sixty-forty. One cash flow. Two listings. The prices wandered apart anyway — by as much as a third — each drifting toward the mood of the market it traded in. A hedge fund run by Nobel laureates bet more than two billion dollars that the gap would close. It widened instead, from eight percent to twenty-two, while the fund died of larger wounds.

That is what happens when the same claim trades in two places and something blocks the trade between them. The gap is not a glitch. It is the rule. Tokenization is about to raise a fresh wall, and put a fast, global, always-on market of copies on the far side of it.

* * *

Most days the gap will be a rounding error, and no one will care. The danger is the day it isn’t.

In March of 2020, the dullest assets on earth came apart. Investment-grade bond funds — the kind your pension holds — traded five percent below the value of the bonds inside them. For a few days there were two numbers. Run it on your own screen: the statement says one, the bid says another, and you must decide, today, with your money, which one is real. The gap closed only when the Federal Reserve promised to step in — before it had bought a single bond.

That was the gentle version. A working democracy, a rescuer standing by.

The hard version came in 2022. Investors around the world held billions of dollars of receipts for Russian shares — a Western wrapper around a Moscow-listed stock. Sanctions on one side and Moscow’s counter-decrees on the other cut the bridge overnight. Holders of the copy were left with paper worth nothing, or with blocked shares they could not sell and dividends they could not touch. The representation did not fall. It ceased to exist.

That is a token with the good years stripped off. A claim on a bridge, and the bridge is only as sound as whoever holds it.

* * *

Look again at what you are, holding the copy. Not an owner with a claim on an asset. A creditor with a claim on a company, or a holder of a token that answers to a list. If the thing behind your copy fails — the platform, the bridge, the rule turned against you — there is no share in the vault with your name on it. The protections built for the original were never stretched to cover the copy.

And when the token and the real entitlement finally disagree, the letter has already chosen. The truth is not on the blockchain: the official record lives in the institution’s own system, and that record “would constitute DTC’s official books and records.” If chain and record differ, the record wins. And behind the record sits the master key — a “root wallet” whose keys can “convert, transfer, mint, or burn any of the Tokens, even without the private key for the Registered Wallet.” You are the last name on a list you were never shown.

You need not take this from me. Wall Street’s own trade group told the regulators what a market of mismatched copies would mean: “multiple prices for a stock in different forms on different markets,” a divergence that would “diminish the strength of U.S. capital markets.” The people who build these markets can see the split coming. They said so on the record.

* * *

Let me be precise, so no one can wave it away.

It does not mean your stock is fake. On the regulated exchange, the token and the share are the same, and the law still holds them together. It does not mean a plot. The motive is ordinary — speed, reach, a new thing to sell.

What it means is narrower, and it is enough. Beside the real market, a second market of copies is being built — faster, cheaper, easier to reach, and not the thing. Most people will be poured into it. And the copy in their hands is programmable: it checks a list before it obeys, and answers to code instead of to them.

They will be right to trust it, most of the time. That is exactly what makes it dangerous. A copy that behaves for years teaches you to stop asking what it is.

So ask now, while asking is free. Ask your broker, in writing: which do I hold — the share, or the copy? The machine turns on in July of 2026. The full service arrives in October.

On the day the two come apart — a frozen market, a severed bridge, a rule rewritten at the center — which one will be in your hand? The share in the vault, with every right attached, that answers to law? Or the quick, weightless copy that was built for you to hold, that answers to code, and was never the thing itself?

Please be careful.

* * *

Every box, its own lock. Now one key opens them all. — Tim Evans / Unsplash

The Permission Society · Part III

On the fourth of May, a company posted a press release.

It was a few hundred words — a corporate notice that scrolls past a thousand inboxes, read by almost no one. Something about a working group, and a launch planned for later in the year.

The company was DTCC. You have probably never thought about it. But it is the quiet center of American finance — the room where the records are kept for nearly every share and bond in the country.

If you own a stock, you do not really hold it. A firm called Cede & Co. holds it, on behalf of a chain of middlemen, on behalf of you. DTCC is where that ledger lives. It keeps custody of more than a hundred and fourteen trillion dollars, and settles nearly four quadrillion in transactions every year.

That is the institution that posted the press release.

And what it announced, beneath the language of modernization, was this: the most liquid assets in America — the largest stocks, the index funds, the Treasury bonds — will be turned into tokens that can be reversed, burned, or stopped from moving, from the center.

No hearing, no vote in Congress, no public rule that anyone could comment on. Just a vendor notice, posted on a Monday, fitting a master key over the most liquid securities in the country.

This is the third piece in this series, and it is the one I most want you to sit with. Because it is the least dramatic, and the most important.

* * *

The first essay was about the token — the thing you own that is really an entry, with an author. The second was about the money — the dollar that asks permission before it moves. This one goes underneath both, to the machine itself.

* * *

The builders deserve a real hearing. Most of their case is right, so let me make it as strongly as I can.

This is not a crypto stunt. It is not a memecoin. The institutions behind it are the most serious in finance: BlackRock, J.P. Morgan, Citi, Goldman, Schwab, the clearing firms, the exchanges. Fifty of them, in a working group. A sitting SEC commissioner welcomed it, calling it a step toward moving markets onchain.

And their central promise is true. Tokenizing your shares does not strip your rights. Under the law, you keep what you had — the dividend, the vote, the claim in a bankruptcy. The securities stay registered to Cede & Co., exactly as before.

The reason is real too. The old plumbing is slow — trades take a full day, billions in collateral trapped in the wait. The new rails settle in seconds, around the clock. That is not a toy. It is a large, real efficiency, and the people chasing it are not fools.

I spent forty years inside this system. I know the people who build these things. They are not cynics. They are engineers who look at a slow, breakable machine and want to make it fast and sound. That instinct is honest, and the problem is real. Hold that, because the rest of this depends on you believing I mean it.

* * *

To understand what is changing, you have to see how little you already hold.

Since the paperwork crisis of the late nineteen-sixties — when Wall Street drowned in physical certificates and firms failed under the weight — the paper was locked in one vault, and only the records moved. That vault is the Depository Trust Company. The shares were registered under a single name, Cede & Co., a nominee that holds them for everyone.

So you have not really held your shares in fifty years. You hold a claim against your broker, who holds a claim against DTC. What you call ownership is already a record in someone else’s book. Tokenization is not the first abstraction; it is the next one — another layer, another keyholder, between you and the thing itself.

* * *

Now the turn.

The defenders say: same rights. They are correct.

But ownership in the real world is not only a matter of legal right. It is also a matter of friction.

In the old system, freezing your assets took work. A court order, served on a broker, passed down a chain of people — each a step where it could be questioned or refused. That friction was a protection. It meant no one could reach your money instantly, or without someone else knowing.

The new system removes the friction.

The legal right stays on top, untouched. Underneath it sits a control surface that did not exist before. The freeze that used to take a court and a week becomes a line of code that runs in a second.

And there is no one to argue with. The old freeze had a human in it somewhere — a clerk, a judge, someone who could be petitioned. A frozen token has no one. Only the key, whoever holds it, and the silence on the far side of a function call.

Same rights, on paper. A master key, underneath. Both true at once. That is the catch.

* * *

I am not guessing. This is in the federal document behind it — the no-action letter the SEC’s staff issued to DTC on the eleventh of December, 2025. You can read it. I have.

Start with what a key is supposed to mean.

The whole promise of digital ownership rests on one idea. Each asset has a key — a long, secret string only the owner holds. Hold the key, and the asset is yours; no one can move it without you. The industry built a slogan on it: your keys, your coins. That was the entire point.

Now read what the federal document gives DTC.

It grants DTC a “root wallet” on each blockchain, with a key that can “convert, transfer, mint, or burn” the tokens — and here is the line that matters — “even without the private key for the Registered Wallet.”

A key that works without your key. A key above every other key.

The one promise — that only the holder can move the asset — is quietly cancelled. The institution can move or destroy the token whether you agree or not. There is a master key now, and the institution keeps it.

And note where it sits. Not at your brokerage app, but a layer above — at the depository, where your broker’s entitlement is recorded. The layer you can never reach is the one with the override.

I am not surprised that it exists. I am only surprised that they wrote it down.

I have spent years studying how control is built into systems — in the defense and security world, and across the markets where I spent my career. Somewhere in every system that matters, its makers leave a way in. A master switch at the center, justified by safety, held for the day they decide they need it. They almost never put it in writing. Here, for once, it is in plain federal text. The technology changes. The master key does not.

The tokens can travel only between wallets the institution has approved. Each of those wallets is screened against the sanctions list before it can hold anything. And when the institution decides a transfer must be undone — the document has a name for it, a “Condition Requiring Reversal” — the root key reaches in and reverses it.

There is one more line you should know. The official record of who owns what does not live on the blockchain. It lives off-chain, in a system called LedgerScan. The letter says it directly: “LedgerScan’s record would constitute DTC’s official books and records.”

Sit with that. The public ledger, the one everyone can see, is not the truth. The truth is a private database the institution keeps. If the two ever disagree, the private one wins.

So the blockchain here is not the open, tamper-proof thing it was sold as. It is a fast messaging layer beneath a central ledger, governed by a master key. Every freedom the technology promised has been quietly removed. The control it makes possible was kept, just as quietly.

* * *

And this is not hypothetical. The power to freeze ordinary people has already been used, in sober democracies, more than once.

In 2013, Cyprus ran short of money. Over a single weekend the government reached into private bank accounts and took a share of what it found — close to half, for the largest depositors at one bank. People woke to find their savings levied and the cash machines capped — only a deal struck while they slept.

In 2022, Canada invoked emergency powers and let banks freeze accounts with no court order. Around two hundred were frozen, for people tied to a protest the government had declared illegal. Whatever you thought of the protest, notice the mechanism. No charge, no conviction — a name on a list, and the money stopped.

That was the slow version — account by account, bank by bank. What the letter describes is the fast version, built into the asset itself.

And the stakes are not only America’s. Foreign investors hold some nine trillion dollars in US Treasuries and twenty trillion in American stocks; DTC holds securities from over a hundred and fifty countries. A growing share of the world’s safe assets is being moved behind one key. That is not a charge against America, but a measure of how much rests on getting the checks right.

* * *

Now consider how it was done.

A change this large — rewiring the ownership records of the country — would normally need a formal rule. A public filing, open for comment. Months in daylight, where anyone could object.

It did not go that way. It went through a no-action letter — the staff saying they will not recommend enforcement, so long as DTC behaves as promised. No public process. None of the rule changes that normally govern critical market infrastructure. And it is temporary: it expires three years after launch, and the staff can revoke it whenever they choose. The people whose ownership was re-plumbed were never asked, because the path chosen did not require it.

And the intent was not sinister. This was no back-room conspiracy. It answers a policy directive — a 2025 federal working-group report urging regulators to make room for exactly this. It is policy, pursued in the open, by people who believe it is progress. And it is not the work of one party. Administrations of every stripe reach for power this way, and each builds tools the next inherits.

But look at what it produced. The most consequential change to American financial plumbing in a generation, installed with no public vote, on a temporary permission, by an agency’s staff. The road it took around the public should give you pause.

* * *

And this power has real work to do. Criminals launder money. Networks finance violence. Sanctioned regimes hide their wealth. A system that can freeze a thief or a terrorist’s account in seconds is not a dystopia. It is a legitimate tool, and I would not wish it away.

A century ago, watching a new technology meet old power, Justice Brandeis named the danger. “The greatest dangers to liberty,” he wrote, “lurk in insidious encroachment by men of zeal, well-meaning but without understanding.” Not villains. Men of zeal, building well. That is what this is.

So the question is never whether the power should exist. It is where it should sit, how fast it should act, and what stands in its way.

* * *

So here is where I land, after all the fairness I can give it.

The rights are intact, and the efficiency is real. And I still think this is a grave mistake — one I believe we will come to regret.

We have taken the records of nearly all American wealth and fitted them with a single master key. We did it for good reasons — to settle faster, to stop fraud, to obey the law when it knocked. Every one of those reasons is sound. Not one of them changes what was built.

Every other great power in a free society is checked. The president by Congress, Congress by the courts, the courts by the people. We built a country on the idea that no hand should hold too much, unwatched. This power slipped past that idea. And the brakes that restrain everything else — a vote, a hearing, a court, an appeal — were left off.

Because a master key does not care why it was made. It knows only what it can do, and whose hand it answers to.

* * *

Let me say exactly what this proves, and what it does not.

It does not mean there is a plot. The motive is public and rational — the old system is slow, and Wall Street wants it fixed.

Nor does it mean your stocks are being seized, or screened for your politics, or tied to a social score. There is no evidence of any of that. The checks today are sanctions and identity. Nothing more.

The key is not silent, either. Each time the institution uses it, it must report that use to the regulators. For now, it is watched.

What it means is narrower, and I think it is enough. The machinery to freeze, reverse, and erase any of these assets — instantly, from the center, by one hand — has now been built, approved, and is being switched on, at the root of American finance. It was built to stop fraud. It will work on anything. It waits only for a reason.

And one day it will not be someone else. The day the key turns on your account, you will not be a criminal or a dissident. You will be a name that reached a list you never saw, for a reason no one is required to explain. By the time you think to ask, the money will already be still.

* * *

So go back to the press release — the boring one, the few hundred words almost no one read.

That is how it happens. Not with a crisis, not with a vote, but with a Monday notice about a software upgrade, and a public trained to scroll past exactly this.

The most liquid wealth in America, and much of the world’s savings parked beside it, now has a master key. You will never see it on your screen. It was cut for the best of reasons, and it will outlast every one of them.

The question is not whether the key will be used well. For now, it probably will be.

The question is the one to ask while the permission is still temporary and the rule has not yet set. Who holds the hand that turns it — and what will we do on the day the reason to turn it changes?

Cash clears without asking. That is the thing we are about to lose.

The Permission Society · Part II

In a few weeks this spring, two pots of money were frozen.

One belonged to a sanctioned government. The world was divided.

The other belonged to a woman with a small online shop. She had done nothing, and no one noticed.

The same power switched off both. It does not ask who you are.

Start with her. She runs the shop alone, and she had just had her best stretch of sales in years. On Monday she opened her account to pay her suppliers, and the money was gone. Frozen.

There was no theft and no charge. Only a form email about unusual activity, and a balance she could no longer touch. Tens of thousands of dollars, held for a hundred and eighty days. Some she would never see again. She broke no law. Thousands like her are now names on a class-action suit.

We have learned to shrug at this. A processor froze the money; call them, wait, hope. Somebody else’s bad luck.

Hold it next to the oldest object in your wallet.

There is a paper note in your pocket that asks no one’s permission.

Hand it to a stranger for a coffee. The coffee is bought. No one is consulted. No server checks a list. No company approves the transfer. The note changed hands, and that was the whole of it.

This is so ordinary that we have stopped seeing it. A dollar bill is one of the last objects in your life that does what you tell it, the moment you tell it, without asking.

That is ending. Quietly, and mostly with our consent.

In March of this year, a company called Circle received an order from a court in New York. The order was sealed; the public could not read it. Acting on it, Circle reached into the digital dollars it issues and froze sixteen business accounts at once.

The businesses were not cartels. They were exchanges, payment firms, a foundation’s software contract. No one had been convicted of anything. A judge signed a paper, the paper named some wallets, and the code did the rest. Several of the frozen had broken no law. A few were quietly thawed later, after the complaints came in.

Hold those two pictures side by side. The bill that asks no one. And the dollar that can be switched off by a sealed order, from a desk, for businesses that did nothing.

That is the subject of this essay. Not crypto. Not the price of anything. Money itself, and what it is becoming.

I ended the last piece on a line: money is the rail everything else rides on, and the first thing they came for. This is why.

* * *

Let me be plain about the thing itself.

A stablecoin is a digital dollar. A private company takes real dollars, holds them in reserve, and issues you a token meant to be worth a dollar, always. You can send it anywhere, instantly, day or night. In most ways it is a better dollar.

It is also a programmable one. That word is the whole story.

A paper dollar is dumb. It cannot be told what to do. It has no memory, no rules, no opinion about who holds it or why. That dumbness is its freedom.

A programmable dollar is not dumb. It is a small piece of code. Before it moves, it checks. Is this sender allowed? Is this receiver allowed? Only if the answer is yes does the money go.

Most days, for most people, the answer is yes, and you never see the question being asked. But the question is always being asked. That is the change. The dollar has gone from a thing that obeys to a thing that asks.

Call it what it is. Permission money.

The biggest of these dollars has a function in its code named destroyBlackFunds. Remember the name. We will come to what it does.

* * *

Now let me give the other side its due, because it is owed, and because much of it is true.

The old way of moving money is bad, and I say that as a man who spent forty years inside it. A wire crosses a border by crawling through a chain of banks, each taking a cut, each closed on weekends. To send two hundred dollars home, a worker pays more than six percent on average. Through a bank, the toll climbs to fifteen. The poor pay the most to move the least. That is not a small injustice. It is a tax on the people who can least afford one.

The digital dollar erases it. It settles in seconds, for a fraction of a cent, at three in the morning on a Sunday. A man in Lagos can be paid by a client in London before either of them finishes their tea.

This is no longer a fringe experiment. Visa is now settling around seven billion dollars a year across these rails, quietly, for ordinary merchants. And the next users may not even be people. Software has begun to pay software, with no human pressing a button. Whatever money is about to become, it is being built right here, in this object that asks permission to move.

And where the local money is dying, the digital dollar is a lifeline, not a toy. In Argentina, the peso lost most of its value, and the government rationed access to real dollars. So families keep their savings in digital dollars, bought in cash-filled back rooms. In Turkey, after the lira fell, the digital dollar became one of the most-held assets ordinary savers reach for. These are not speculators. They are people trying to keep their wages from melting in their hands.

And it even serves the state. Every digital dollar is backed by Treasuries, so Washington wants this too. The state, the companies, and the man in Lagos all want the same thing. That is what makes it so hard to stop.

I want to be clear, because the rest depends on it. This is real good, done for ordinary people. The men building it are not, for the most part, villains. If the digital dollar were only a trap, it would be easy to refuse. It is not only a trap. That is exactly the difficulty.

This is the first room. I have sat in it. The people there are clever, and mostly sincere, and often right.

Even its critics mostly worry about the wrong danger. They warn it could collapse: a run, a panic, a bailout. Maybe. But that is the old kind of danger, the kind we know how to name. Mine is quieter. It is not that the money breaks. It is that it works perfectly, and does exactly what it is told.

* * *

Now the turn.

The paper dollar in your pocket is a bearer thing. Possession is the whole of ownership. Hand it over and the matter is closed.

The digital dollar is not a bearer thing. It is a claim. You do not hold a dollar; you hold a promise from a company that it will give you one. And the promise lives on a ledger the company controls.

How tightly does it control it? Of all the people and firms holding these digital dollars, only a select group of institutions has a direct line to the issuer. The right to walk up and redeem at face value. Ordinary holders do not have it. With Circle, an individual cannot become a direct customer at all. Everyone else holds a claim on a claim. A second-hand promise, passed from hand to hand, redeemable directly by almost no one.

Set the two side by side. Cash: no one can stop it, final on delivery, with no one to ask. A digital dollar: the issuer can stop it, not at your account but at the level of the money itself. A frozen digital dollar is frozen in every wallet on earth at once.

So when the dollar asks permission before it moves, whose permission is it asking? Not yours. The issuer’s. And behind the issuer, the state.

This is not a figure of speech. It is the literal order of operations. Every time one of these dollars moves, the code runs a check first. Is this address blocked? Only then does it release the money. The transfer you picture as handing over cash is, underneath, a query to a private database. Asked and answered before anything happens.

The dollar consults a list about you before it agrees to leave your hand.

* * *

This is not hidden. It is written into the law, in the law’s own words.

Last year the United States passed its first federal statute for these digital dollars, the GENIUS Act, signed in July. Most of it is sensible plumbing — reserves, audits, disclosure. Set into it is a requirement that every issuer be able to obey what the law calls a “lawful order.”

A lawful order, the statute says, is one that requires a company to “seize, freeze, burn, or prevent the transfer” of its dollars.

Seize. Freeze. Burn. That is the actual text. The power to reach into your money and end it is not some dark feature a rogue engineer might one day add. It is a condition of the license. You cannot legally issue a digital dollar in America unless you can switch it off.

And the switch is real, down to the line of code. Now, destroyBlackFunds. It is Tether’s, and it does what it says. It takes an address, sets the balance to zero, and erases the money from existence. No private key required. No bailiff at the door. A command, run on a public network, in the time it takes to read this sentence.

We have watched it used at scale. In April, at the Treasury’s request, Tether froze three hundred and forty-four million dollars in accounts tied to Iran. It was done in a day. In the old world, freezing a hostile state’s money meant months of letters between central banks. Now it is a function call.

When the target is a sanctioned government, the freeze is easy to accept. Hold that thought. We will need it.

The reach is also widening. A new Treasury rule this spring would widen the net. Issuers would screen not just the customers they deal with directly, but transfers between strangers, far out in the open market. The permission layer is not meant to sit at the door of the bank. It is meant to travel with the money, wherever it goes.

There is a second move here, quieter than the freezing, and in the long run more important. The cash itself is being pushed out. In 2023, Nigeria’s central bank pulled most of the country’s banknotes from circulation almost overnight, trying to herd its people onto digital money. For weeks, ordinary Nigerians could not buy food or fuel. The cash was gone, and the digital rails buckled under the weight. It was sold as modernization. What it showed was simpler. Once the dumb, permissionless option is taken away, you are left only with the kind of money that asks. The freezing is the visible danger. The quiet retirement of cash is what will make the freezing matter.

Notice the speed. The law was signed in July. The code was already written. By spring the freezes had begun. This did not take a generation. It took a year.

* * *

Here is where I land, after all the fairness I can muster.

The good is real, and I would still not make this trade.

We are taking the one ordinary object that did what we told it, and replacing it with one that does what it is told. We are doing it because the new one is faster and cheaper, and it is. In exchange, we accept a switch into the base layer of daily life. And we hand that switch to a few private companies, and to the government that licenses them.

I do not think most of the people building this intend tyranny. I have sat in the rooms where these things are drawn on whiteboards. They intend efficiency. They intend to stop criminals and sanctioned regimes, and they will, and we will applaud.

But a switch does not care why it was built. It knows only the rule it is given, and the hand that will write the next rule. We are building it to stop the worst of us. It will work just as well on the rest of us. That is not a claim about anyone’s character. It is a fact about the machine.

The day it happens to you, you will not be a criminal or a dissident. You will be someone whose rule changed overnight.

You have met a smaller version of this power — the woman whose shop money was frozen, the creator cut off overnight. We learned to shrug. The programmable dollar takes that scattered, ad-hoc power, the bank’s and the processor’s, and builds it into the money itself. Everywhere, switched on by default. That is permission money: the freeze she suffered once, by accident of policy, made total and automatic and built into the coin.

There is a turn here that is easy to miss, and it cuts against what most people expect. We were told the danger was a government digital dollar, the state watching every purchase from its own ledger. America has, for now, turned that down; a central bank digital currency is barred by executive order, and the House has voted to ban it outright. We congratulated ourselves.

But look at what we built instead. The same switch, the same freezing and burning, the same surveillance of the flow of money. Only now it sits inside private companies, acting on the state’s orders. A government ledger would at least be bound by the Constitution, by the protection against unreasonable seizure and the right to due process. A private company is bound only by its terms of service. And those terms reserve the right to freeze your money, in their own words, at their sole discretion. We did not refuse the surveillance state. We outsourced it, and stripped it of its constitutional restraints on the way out.

Tocqueville saw the shape of this, in a gentler century. He warned of a power that would not break men’s wills so much as soften them. An orderly, provident authority. One that people accept willingly, because it spares them the burden of choosing for themselves. A servitude that arrives wearing the face of convenience. That is what this is. No one will seize our money from us. We will hand over the right to spend it freely, for speed and a few cents saved, and we will be grateful for the service.

And ours is the mild version.

Tocqueville’s soft despotism belongs to the comfortable democracies, where a court still has to sign the order and a constitution still, on paper, draws a line. Most of the world has no such brakes. The same rail is being laid in countries where the courts answer to the ruler and the constitution is a decoration. There, the switch will not be saved for foreign regimes under sanction. It will be turned on the dissident, the protester, the minority, the inconvenient. The code does not change at the border. Only the restraints do, and across much of the world there are none.

And here is the bitter part. The people the digital dollar rescues from their own broken money — the family in Buenos Aires, the worker in Lagos — are the same people left most exposed when the switch is turned the other way. We are building the most precise instrument of financial control in history, and handing it, rail and all, to governments that have never once been told no.

* * *

Let me say plainly what this does not mean, so no one can wave the argument away.

It does not mean every freeze so far has been wrong. Stopping an outright thief is easy to defend. But a sanctions list is written by a government. Who goes on it, and who decides, is its own political fight. The power to freeze is only ever as just as the hand that writes the list.

It does not mean there is a plot. There is no need for one. Each step is useful, legal, and defensible on its own. That is precisely why no one will stop.

What it does mean is narrow, and I think it is enough. The power to freeze, seize, and erase any dollar is now standard, legal, and in routine use. It was built for the guilty. It is available against everyone.

* * *

So go back to the note in your pocket. The dumb one. The one that asks no one.

Soon it will be the strange one, the relic your grandchildren find quaint, the way we find a gold coin quaint. The dollar they spend will be quick and clean and clever, and it will ask permission every time it moves, and most of the time the answer will be yes.

The question is not whether to use the new dollar. You will. So will I.

The question is the one to ask now, while we still remember what it felt like to spend money that answered to no one. Ask it while the asking still costs you nothing.

When the dollar can say no, who decides what it refuses — and what will you do on the day it refuses you?

A boarded house, owned now by people who will never see it. — Liam Rearick / Unsplash

The Permission Society · Part I

Somewhere in Europe, at three in the morning, a phone lights up.

Its owner has earned a few cents on a house in Detroit. The number in his app has ticked up while he slept.

At that same hour, in that same house, Cornell Dorris is awake and standing in water. It is black, and it comes up through the basement floor when it rains. Tonight it is raining.

Dorris does not own the house. About sixteen thousand people do, scattered across the world. Sometime around 2023 the building was cut into pieces and sold, fifty dollars a slice, as digital tokens on a blockchain. The sleeping man owns one of those slices. He has never seen the basement. He has never met Dorris, who has lived here for ten years, cooks for a living, and knows the place the way you know a house you have slept in for a decade.

A line in a ledger connects them. Nothing else does.

The brochure is real. So is the basement. Both at once. That is the problem this series is about.

Let me give you the ending now, so you can watch me earn it. The thing the sleeping man believes he owns, he holds by permission. The deed to that house sits in a county office in Michigan, and his name is nowhere on it. He owns an entry in a ledger. An entry has an author. And the author is not him.

Hold that. We will come back to it from the basement.

* * *

First, what the thing actually is.

A token is a digital stand-in for something you own. A share. A bond. A dollar. A bar of gold. A house.

The thing itself does not move. The gold stays in the vault. The house stays on its lot. What moves is the record of who owns it. The record goes onto a blockchain. A shared ledger, read by many at once, instead of a private one kept in the dark by a single firm.

That is all a token is, at the bottom of it. A line that says this is yours.

There is a fork here worth marking, because it matters later. Some tokens are the thing itself. Some are a claim on a thing that someone else holds for you. And some are neither. They only copy a price, the way a photograph copies a face. On the screen, all three look the same. The sameness is the point.

It sounds like a filing change. A faster database. By the end of this you will see why it is not.

* * *

Now the other side, in full, and without a sneer.

I have spent my life around capital and the machines that move it. So I know what the old plumbing costs. It is slow. It is dear. It drops things.

Buy a share of stock today and you do not get the share. You get an entry at your broker. Your broker holds an entry at a clearing firm. The clearing firm holds an entry at one company that keeps nearly all the stock in the country under a single name. The trade takes a day to settle. Sometimes two. Behind the screen, people spend their working lives checking one list against another.

Tokenization promises to fold that whole tower into a single page. One ledger. Settlement in seconds. A market that never sleeps.

This is not a dream. Pieces of it already run.

On the sixth of May, 2026, a block of United States Treasury bills was redeemed across a public blockchain. The asset cleared in under five seconds. The cash ran through a bank network and reached an account in Singapore minutes later. The clearinghouses were closed. The money did not care.

Five seconds. Hold that against the basement. In Detroit the water takes all night to drain. In Singapore the safest asset on earth moved before you could finish reading this sentence.

That speed is no trick, and it is worth wanting. So are the savings, and they have been measured, not guessed. Hong Kong’s central bank studied tokenized bonds against ordinary ones. The tokenized bonds traded tighter. Investors took a little less yield to hold them. The cost of issuing fell by almost a quarter. Those are basis points, and basis points are real money to the people who pay them.

The largest money manager on earth has staked its name on this. Larry Fink runs BlackRock. He has said it out loud, from the stage, more than once. Every stock, every bond, on one general ledger. He is not a man in a basement with a theory. He holds the retirement savings of teachers and firefighters in his hands. When he says it, the room leans in.

And there is the fairness case, which is the one I cannot wave away.

Half the planet carries a bank in its pocket now. A phone. Many of those people cannot open a brokerage account. The old system was not built for them. A token does not ask where they live. If they have a wallet, they can hold a slice of a bond issued by a government an ocean away. For a young man in Lagos, that is a door that was bolted his whole life.

So let me say it, and mean it. The people building this are not villains. The benefits are not lies. If this were only a swindle, it would be easy to dismiss, and I would not be writing about it. It is not only a swindle. That is exactly why it matters.

This is the first room. I have sat in it. Let me tell you about a time I did.

I will not name the city. It does not matter. There have been several, and they are the same room.

A young man stood at the front. Clever. Sincere, which is the part people forget to fear. He believed every word. He had a diagram, and the diagram was beautiful. Boxes, arrows, and at the center one clean ledger where everything met.

He said the word frictionless. He said it the way a priest says grace.

I watched the table. Men I had known for thirty years. Careful men. They were nodding. So was I.

Because it was true. That is what I need you to understand. Every line of it was true. Faster, cheaper, open to the poor, safer even, in the narrow way he meant it. I had spent forty years pushing money through pipes that leaked. Here was a man with pipes that did not.

I wanted it. I will say that plainly, because there is no use in any of this otherwise. I did not sit there as a prophet seeing a cage. I sat there as a buyer, wanting the thing on the screen.

Then he reached the slide about compliance. The keys. He said the issuer would keep certain controls. For safety. For recovery. For the law. He said it quickly, in a smaller font, and moved on. No one asked. I did not ask.

I have thought about that silence for a long time.

We were not deceived. He hid nothing. The power to reach into the asset and switch it off was right there, in the smaller font, and we let our eyes pass over it, because the rest of the picture was so clean, and we wanted the clean picture.

That is how it comes. Not with a gun. With a beautiful diagram, a smaller font, and a room of careful men who want the convenience and do not ask about the key.

I was one of them. That is the part I cannot set down.

* * *

Now the turn.

Go back to the deed.

When I sold my land in Montana, what passed was a piece of paper. It sat in a county office. It said what it said. No one could rewrite it from a thousand miles away. To take it, a person would have had to stand in that office, or send a sheriff, or win in a court of law. The paper was slow. That slowness was its whole virtue.

A token is not a piece of paper. A token is an entry. An entry has an author. The author holds a key. And the key can reach in.

This is the seam. It is easy to miss, because the pitch is built to slide you past it.

Here is what I mean.

When a firm like BlackRock puts a fund on a blockchain, the token for your share is not a wild, free coin. It is built to be governed. It has to be.

The code has a privileged role written into it. Call it the agent. The agent holds a master key. With that key, the agent can freeze your token where it sits. Move it out of your wallet, with no signature from you. Destroy it, and mint a fresh one for someone else.

This is not a flaw. It is the feature. It is written down, in the open, in the standards these tokens are built on. One of the most common says it without blushing. The issuer always keeps control. Mint, burn, block, or force transfers, at any time.

At any time.

I am not describing a break-in. I am describing the blueprint. The power to reach into your holding is not a defect the engineers missed. It is the product they shipped.

And there is a reason for it. A decent one. Lose your key, and someone must be able to give your money back. A court orders a freeze, and someone must obey. A thief drains you, and someone must claw it back.

Each of those is a service. Each is the same power, turned a different way.

The power to give your money back is the power to take it away. There is no version of the first without the second.

I know how that sounds. I also know I would have wanted that recovery key the morning I lost my own. I would have called it protection. I would have been right. That is the trap. The thing that saves you is the thing that holds you.

* * *

And here is the part that was true long before the blockchain arrived.

You think you own your shares. You do not. Not the way you own a chair in your kitchen.

Under the law, what you hold is a claim. A claim against your broker. Your broker holds a claim against the firm above it. At the very top sits one name, holding almost all the country’s stock for everyone at once.

A claim against a claim against a claim. A hall of mirrors. Your name is in none of them.

You have been a tenant in another firm’s ledger for fifty years. You simply never had to look.

Tokenization did not invent this. It inherited it. Then it bolted a faster, finer set of controls on top. The token does not free you from the middleman. It hands the middleman a sharper tool.

Faster to settle. Easier to audit. Also easier to freeze. The same upgrade does all three at once, and you do not get to choose which one you are buying.

* * *

There is a second move, quieter than the first.

Some tokens hold the real thing. Behind them sits an actual share, carrying the rights a share carries. Others hold nothing of the kind. They track a price and grant you nothing underneath. No vote. No dividend. No claim on the company at all.

In January of 2026, the staff of the Securities and Exchange Commission wrote the difference down. Some tokens carry the rights beneath them. Others may or may not. In that second case, what you own is a promise from whoever minted the token. It runs to the middleman, not to the company.

Europe’s market regulator said it more bluntly. A tokenized stock can follow the share price and give you none of the ownership. The danger, it said, is that people will not understand. They believe they bought the thing. They bought a picture of it. The world’s stock exchanges, no friends of crypto, sent the same warning in 2025, and asked the plain question no one could answer. If the platform fails, what does the holder actually own?

* * *

Now back to the house.

This is where the brochure tears open, and it tears along the line I drew at the start.

American law will not let you sell land with a token. It cannot be done, and the wall is deliberate. Land passes by deed, recorded in a county office, and the law guards that road on purpose. There is a doctrine, old as English law, that says you cannot invent a new kind of property and turn it loose in the world. The forms are fixed. The registry is the truth.

So the people who tokenize houses do something clever. They do not tokenize the house. They cannot. They put the house inside a company. A small shell, one for each property. Then they sell you tokens in the shell.

You do not own a piece of the house. You own a piece of the company that owns the house. Your name is not on the deed. It never was.

When it works, you cannot feel the gap. The rent lands. The number rises. You feel like an owner, and the feeling is enough.

When it breaks, the gap is the only thing left.

In Detroit, it broke. The city sued the company that sold those tokens. It named the two founders and a hundred and sixty-five of their shells in a single filing. It said hundreds of the houses were unsafe. Mold. No heat. No certificate fit to be lived in. The largest suit of its kind the city had ever filed.

A judge stepped in. She turned the rent away from the wallets and into an escrow account, to be spent on repairs. The sleeping man stopped being paid. The water in the basement got a budget.

I will be careful here, because it would be easy to overreach, and I will not.

What happened in Detroit is not the dark future I am warning about. It is something older and plainer. Bad landlording. Slumlords were here long before blockchains. You do not need a token to let a roof rot.

But Detroit shows the seam, lit from the inside. The investors believed they owned homes. They owned shares in shells. The ledger said one thing. The deed said another. When the two disagreed, the people in the houses paid for it, and so did the people in the wallets.

The token felt like ownership. It was a claim, sitting on a claim, sitting on a deed that none of the owners ever touched.

As for the company, it said it would sell the Detroit houses and move on. New tokens. New buildings. Colombia next, then Panama. The ledger rolls forward. The basement stays where it is.

* * *

Now some cold numbers, because the talk runs hot and the truth is smaller, and stranger, than either side will tell you.

Set the digital dollars aside for a moment. The whole world of tokenized real assets was worth about six billion dollars at the start of 2025. By the middle of 2026, around thirty billion. Fast growth. Also tiny. The American stock market alone runs to the tens of trillions.

And most of that thirty billion is not houses or art. It is the unglamorous end of finance. Short-term government debt. Private credit. Some gold. The standardized stuff moved first, because that is where the back-office savings are real.

The digital dollars are the bigger story. Stablecoins are now worth around three hundred billion. The money layer dwarfs the asset layer. Remember that. Money is the rail everything else rides on, and money is the first thing they came for. That is the next essay.

And the small investor’s dream, the slice of a house? One study looked at fifty-eight tokenized rental homes. Each had, on average, two hundred and fifty-four owners.

Two hundred and fifty-four people to a single house. None on the deed. None in the basement.

* * *

So let me be exact about what this proves. And about what it does not.

It proves the old plumbing can be replaced, and that the new version is faster and cheaper at what it is built for. Cash. Collateral. Government debt. That is here, and it is working.

It proves idle money can be made to work around the clock. A true gain. The people chasing it are not fools.

It does not prove that tokenizing a thing makes anyone want it. A house nobody wanted is still a house nobody wants, only now in slices. Even the builders admit it. You cannot conjure a buyer with code.

It does not prove that anyone slipped the law. The regulators did not vanish. They wrote themselves into the system. There is even an argument from inside the government that the slow settlement everyone wants to kill is a feature, not a fault. The daily netting it allows cancels about ninety-eight cents of every dollar before any money moves. Speed is never free. It only hides the cost.

And it does not prove the small investor got what he was sold. Often he got a shadow. A price without a vote. A yield without a deed. A landlord’s cut with none of a landlord’s reach.

* * *

So here is where this leaves us.

Take the speed. Take the convenience. They are worth wanting, and you will want them. I am not telling you to refuse. I am telling you to notice the trade. What you used to hold outright, you now hold by permission. The deed in the drawer answered to no one. The token answers to whoever holds the key.

A deed is a fact. A token is an entry. An entry can be edited.

Dostoevsky drew the shape of this a long time ago. His Grand Inquisitor tells Christ that men will lay their freedom at his feet for bread, and thank him for taking the weight of it away. He understood us better than we like. No one will seize what we own. We will hand it across the counter ourselves, for the speed, for a few cents arriving in the night, and we will be grateful for the service.

Go back, one last time, to the man in Europe. He is asleep, a few cents richer, on a house he will never see. And to Cornell Dorris, awake in the water, in a house that strangers on the far side of the world own and cannot help him fix.

One line in a ledger connects them. Someone, somewhere, can change that line.

So before we go further, ask the small question first. Not whether to use this. You will. We all will.

Ask who can reach in.

Nazem Alkudsi

The Old way of recording who owned what

A piece of ground I used to own sits in Montana, hard against the edge of Glacier. People assumed it was farmland when I said I owned land there. Nothing about it was tidy enough to farm.

It was a place where three worlds met. The Pacific forest came in from the west, dark and close. The alpine tundra came down off the peaks. And the Great Plains ran up against all of it from the east, dry and windblown, held back from the mountains by nothing but a rain shadow. You could stand in one spot and see all three: lodgepole and Douglas fir behind you, mixed-grass prairie ahead, limber pine bent along the ridges where the wind never lets up.

I thought a great deal, standing there, about what it meant to own such a place. I think about it more now that I have sold it.

The lodgepole pine carries its seed in a cone sealed shut with resin. The cone can hang on the branch for years, closed, waiting. It opens only in fire. The tree gives up its future to the one thing that looks like ruin, and trusts the heat to do what no season could.

You cannot put that in a ledger.

When I sold it, the land passed the way land has always passed. A deed signed across a desk. A name struck out and another written in. A stack of paper that meant exactly what it said. It changed hands cleanly, with nothing held back, and no one able to reach in afterward and reverse it. That is one of the last kinds of ownership we have left.

This series is about the deed, not the land. About what it still means to own a thing outright, and to let it go outright, in a world quietly building a system where almost nothing will be owned that way again. And I will tell you plainly, before we begin, where I have landed. I think this is likely to go badly. Not certainly. But likely. Because the world we are building makes taking what people own easier than it has ever been in human history. Not only redefining it, though it does that too. Taking it. Freezing it, filtering it, switching it off, lifting it clean away. The thing that once required an army will soon require a keystroke.

* * *

Within living memory, the government decided that one thing Americans owned outright would no longer be theirs to hold.

In the spring of 1933, with the banks failing and the country frightened, Franklin Roosevelt signed an order forbidding citizens to hoard gold. Bring it to a Federal Reserve bank, take paper money at a price the government set, and do it by the first of May. The penalty for keeping your own gold was a fine large enough to ruin a family and ten years in prison.

It is remembered as a confiscation, and in law it nearly was. The truth is stranger. The order was riddled with exemptions. It was enforced by almost no one. There were no raids on homes. No agents drilling open the nation’s safe-deposit boxes. The lurid accounts of that came later, and they are invented. Milton Friedman and Anna Schwartz later tried to account for the gold. They found most of it had simply never come in. Only a fifth to a quarter was ever surrendered. The rest stayed where it was. Buried in yards. Hidden in floorboards. Kept.

Ordinary people could decline because the gold was physical, a thing you could close your hand around and put somewhere a government could not easily follow. The floorboard was the last refuge of the man who would not comply, and it held.

The rest of this series is, in a sense, about the day the floorboard disappears.

* * *

There is a longer catalogue of governments seizing what their people held, and some of its pages are very dark. I will set most of them aside. They teach the wrong lesson for this moment. The confiscations that came with force were the ones people could see, and flee, and hide from. I want to turn only one page, because it is mine.

My family is from Syria. We had people in Egypt as well, and the same thing reached Iraq. Anyone whose roots run through that part of the world grew up hearing what happened when the socialists came.

They did not call it theft. They called it justice. The redistribution of wealth. The return of what belonged to the people. Under that banner they took land that had passed through a family for generations, the olive terraces and the fields, the deeds that meant exactly what they said. They took the factories. They took the businesses a man had built with his hands and meant to leave to his sons. None of it was stolen, in their telling. It was simply redistributed, for the good of all.

It did not produce the good of all. It ruined the countries that did it. The wealth did not move to the poor. It evaporated, the way wealth does when no one is permitted to own it. And two generations later those nations are still poorer for it.

I heard the stories as a child. Men who learned in a single afternoon that everything their family had carried, from one careful hand to the next, was gone. Some of them did not survive the news. A heart simply stops, when a life’s inheritance is read out as someone else’s policy.

I tell you this so you know I am not theorizing from a comfortable distance. I know what it looks like when a state decides that what you own is now a question of permission. I know it was sold as fairness. I know the people who sold it believed, or said they believed, that they were setting men free.

That is the part worth remembering. The worst of it almost never arrives as theft. It arrives as justice, as efficiency, as progress, as a better arrangement for everyone.

But here is what I have come to see, and what I most want you to sit with. What my family lived through was the hard version. It took a government, and decrees, and enforcement, and men with the authority to come and take. It took time. It was resisted, evaded, survived. Wealth slipped through the cracks of it, because the cracks were everywhere, because taking things by force is clumsy and slow and leaks at every seam.

We are building the version with no cracks.

When your house is a token and your money is programmable and your name is a credential that grants or denies you access to both, the act that once needed an army needs a line of code. No decree. No soldiers. No afternoon of terror at the door. One instruction, applied to one account or to a million at once, from a desk, by someone you will never meet. The confiscation my family survived was expensive and friction-filled and resistible. The one we are building is cheap, frictionless, and complete.

I want to be precise, because this is where the careless lose the argument. No one is building a confiscation machine. They are building a convenience machine, and it is a marvel. But a convenience that can lift everything you own with a keystroke is a confiscation machine the moment anyone decides to use it that way. The capacity is the danger, not the intention. We are assembling the most powerful instrument of dispossession ever made, and we are doing it without once asking who will hold it after the people who built it are gone.

I want to be careful here, because the loudest voices on this subject reach immediately for villains. I do not. No conspiracy is needed for any of it. If the cage gets built, villains will not be the ones who build it. It will go up one reasonable step at a time, adopted freely, sold as freedom, until stepping out is no longer practical and you cannot point to the day anyone chose it. The question was never the technology. It is who holds the keys, and whether the rest of us are awake while those keys are cut.

* * *

This series will not traffic in fear. It also refuses the easy comfort, the one that says this cannot reach us here.

There is an older idea, one the world has nearly forgotten. That we never truly owned anything to begin with. That the land was lent to us, that we are stewards and not masters, that everything passes through our hands on its way to the next pair of hands. I believe something like that. But there is a vast distance between holding a thing in trust, from God, or from time, or from those who will come after, and holding it at the pleasure of a platform that can withdraw it while you sleep. The first is humility. The second is servitude. We are being asked to mistake one for the other.

I sold my land with a deed, cleanly, the old way. A name struck out, another written in, nothing held back. The generation coming after me will not sell theirs the same way. And they may not notice the moment the floor beneath them quietly becomes a permission.

That is the thing I mean to write about. Not whether this world is being built. It is being built. Whether anyone will be awake when it finishes.

Pattaya by night. A fishing village before the war, it filled in 1965 with the engineers building the runway at U-Tapao — and never emptied.

The heat comes off the concrete at U-Tapao in a way you can almost see, a wavering in the air above the strip, and through it a small jet from Kunming settles down and taxis toward the terminal. It is an ordinary afternoon in May. The passengers coming down the stairs are tourists, most of them young, looking at their phones for the cars that will take them to the beaches an hour up the coast. None of them look at the runway. It is only a long grey rectangle, wide enough for two aircraft side by side, thick enough that it has not been rebuilt in sixty years.

It was poured that deep for a reason. The strip was laid in the mid-1960s, by American naval engineers, to carry the heaviest bombers the world had then made. The war those bombers flew to was not Thailand’s. The country had not been invaded, had declared nothing, had sent no armies into the fight when the runway was laid. And yet the war came here all the same, down through this concrete and the six other airfields like it, and it did not leave the country the way it found it.

Near the fence an old man in a faded vest watches the tourists go. Behind him the cranes are already up over the next thing, an airport city the government is raising on this ground, worth nearly three hundred billion baht, its concession running to the year 2076. The strip the war built is becoming the foundation of a resort.

Laos was bombed until its plain of jars filled with iron the size of a fist, the kind that still comes up under a child’s hoe. Cambodia was bombed too, and then turned on itself in a way that has no bottom. Thailand was neither bombed nor occupied. It was changed from the inside, by a war it agreed to host and never had to fight. The other two were ruined by what was done to them. Thailand’s wound is of another kind: it chose.

To see why a Thai government would open its country to a foreign air force, go back to 1962 and stand inside the fear of the soldiers who ran the place. They looked north and saw the thing they had been raised to dread. Laos was coming apart, the war in Vietnam was widening, and the communism that had taken China seemed to be coming down the rivers toward the Mekong. Bangkok did not believe, whatever it said aloud, that it could stand against that alone.

Thailand had an old habit of survival. For more than a century its kings had kept it independent by bending, giving a little to whichever great power leaned hardest, never planting their feet so firmly the country could be knocked down. It was the one kingdom in the region never ruled from Europe. But bending works in a world of several powers leaning from several sides, and in the Cold War one direction counted. The men in Bangkok leaned the whole way into it.

On the sixth of March, 1962, the foreign minister Thanat Khoman and the American secretary of state Dean Rusk signed a joint statement whose careful language carried an enormous meaning. The American promise to defend Thailand, it said, was individual as well as collective, which meant Washington need not wait for its treaty partners to act. The Americans were not cynical in this. They believed, with the conviction of their age, that communism would take the whole region if it were not stopped somewhere, and that Thailand was the somewhere. With that sentence the bases became possible, seven great airfields in time, and the better part of fifty thousand American servicemen on Thai soil. The public was told almost nothing of what they were for. The bombing those airfields launched belongs to the countries next door. What stayed was the door they opened into Thailand itself, and almost everything that remade the country came through it.

What came through the door, at first, looked like a miracle.

It came as roads. In the late 1950s, with American funds, Thailand built the Friendship Highway, the Mittraphap, running from the centre of the country into the dry northeast the capital had half forgotten. It was the first road in the kingdom built to an international standard, and it had a double use: it carried supplies toward the bases, and it bound the poor interior to the rest of the nation for the first time. The deep-water port at Sattahip grew on the same logic, and in 1966 a treaty of amity opened Thailand to American business. The frame of a modern economy was being laid down at speed, and the war paid for the steel.

Then there were the soldiers on leave, hundreds of thousands of them, and an economy grew up to take their dollars. Within a few years Bangkok had hundreds of bars built around the off-duty American. At Pattaya, a fishing village before the war, the man who opened the first hotel remembered it filling, in 1965, not with holidaymakers but with the engineers building the runway at U-Tapao. The village became a resort. Young women came down from the northeast to work the bars and sent their wages home, so that the road the Americans had built carried the women south and the money north.

Three older trades were caught in that same current and changed by it. Prostitution and gambling were not new to Thailand, as old as any port; what the war did was industrialise them, give them dollars and permanence and scale, and turn a margin of the economy into something nearer a pillar. The narcotics were a harder case. The opium of the highlands was older than any American, but the networks that now moved heroin, first to the soldiers who wanted it and then into the cities of Thailand itself, were built to the size of the war. None of this was American design. It was the side effect of magnitude, of forcing that much money and appetite through a small country in a few years. The bill came due later, and Thais paid it: addiction through the towns in the early 1970s, districts that did not close when the soldiers left, a new disease that would travel that same road a decade on, and the money that fattened the men who ran the traffic and the violence with them. The road that brought the wealth brought the rest, on the same wheels.

But the deepest thing that came through the door was not money. It was the army. A frontline ally is armed, and Thailand was armed, year after year, with weapons and training and the quieter machinery of a security state. The men who already held power grew richer on it and surer of themselves, and the institution they commanded grew until it was no longer one power among several but the power from which the others took their permission.

This was not new to the country, but the war made it permanent. The army had first seized the government in 1932, when it ended the absolute monarchy, and had taken power by force a dozen times since. What the alliance did was settle the habit into the bones, fund it, arm it, and hand it a Cold War reason that silenced argument. A general who took the state could now do it as the wall against communism, with a great power behind him. The civilian governments that came and went between the coups governed at the army’s sufferance.

A country cannot take in that much of a war and have it touch only its roads and its banks. A society that agrees to stand on the front line of someone else’s war begins, after a while, to treat its own quarrels as battles in it. The students who wanted a freer country could be called communists, and so could the workers who asked for a better wage and the farmers in the northeast who only wanted to keep their land. The fire Thailand had kept off its borders had come in through the door with everything else, and by the mid-1970s it was near the surface.

There is a field at Thammasat University in Bangkok where, on certain October mornings now, people hold sheets of clear acrylic up against the buildings. Printed on them are photographs from one morning fifty years ago, and a person tilts the sheet until the doorway in the old picture lines up with the doorway still standing, the same stone, so that for a moment the past lies over the present like a second skin. The hardest of these shows this same grass covered with young men stripped to the waist, made to lie face down in rows while armed men stand over them. It was taken here. The ones lying in it were students, and many had only hours to live.

Before dawn on the sixth of October, 1976, the police and the army and the gangs the state had armed came onto this campus. For weeks the students had been named communists in league with Vietnam, plotters against the throne, insulters of the heir to it; a newspaper had run a doctored photograph claiming to show a student dressed as the prince. Two men who had hung posters in the provinces were already dead, beaten and strung from a gate. By sunrise the campus was being fired into with the weapons of war, and the students caught in the open had nowhere to go.

What was done then is hard to set down, and was meant to be. People were shot. People were beaten to death with whatever lay to hand. Some were hanged from the trees along the avenue while a crowd gathered to watch, and a photographer caught a man swinging a folding chair into one of the hanging bodies as the people around him smiled. That photograph won a prize in America the next year. The official count was forty to forty-six; the man who had led the university said it was past a hundred. Against what had happened in Cambodia it was a small number. It was not a small thing.

A well-known monk had given the killers their absolution in advance: to kill a communist, he said, was not demeritorious, because the thing being killed was not truly a man. That evening the army took the government and began the work of forgetting. No one was punished. The killing was scrubbed so thoroughly from the schoolbooks that for years a Thai child could pass through a whole education and never learn it had happened.

One student who lived through it, who spent two years in prison and became a historian, has given the rest of his life to making the country remember. He calls what it lives in not forgetting but its opposite, an unforgetting, the state of being able neither to recall a thing nor to let it go. Another survivor, a first-year student on that field, wrote that every October the sixth an old sadness takes her she cannot account for. The army the bargain had built had turned its weapons on the country’s own young, then filed the morning away as carefully as the graves were filed away in Laos.

The army did not loosen its grip after that. Power was taken by force again in 1991, in 2006, in 2014, each time from an elected government, each time in the name of order. Between the coups the courts learned to do the same work by other means, dissolving parties and removing prime ministers the generals found inconvenient. The most popular electoral movement of the century, built around Thaksin Shinawatra and then his sister and his daughter, was put out of power three times in twenty years, twice by soldiers and once, in 2025, by a court ruling.

In February of 2026 the country voted again, and the winner was a royalist who makes no secret of his closeness to the generals, lifted into office on a wave of nationalist feeling stirred by fighting along the Cambodian border, where Bangkok had handed the army full discretion to answer perceived insults with force. A progressive movement that had run on pulling the military out of politics lost. Sixty years after the runway was poured, the institution the war had fed was still the decisive power in the land. The fire the bargain was meant to keep outside, that had burned at Thammasat, had never gone out. It had only changed what it was willing to be called.

So how should one think of Thailand? Two readings offer themselves, and both hold at once. In the first, Thailand is no victim. It saw the danger, judged that alignment was safer than neutrality or defiance, and leaned the whole way in with its eyes open. It took the bases and the money, built the roads and the resorts, and has prospered since on foundations the war poured. Its leaders were not tricked; they made a calculation, and on its own terms it held. Set beside Cambodia and Laos it can even look like wisdom, the wisdom of a man who reads the weather and builds where the storm will not reach.

In the second, the bargain was paid for by people who were never asked. A small country tied to a giant is never the giant’s equal. The army the alliance fattened was turned not outward at the giant’s enemies but inward, at Thailand’s own. The students shot down on their grass had not signed the agreement of 1962. Thammasat was the bill for a choice a few frightened men had made long before most of the dead were old enough to vote.

Why did Thailand bend without breaking when the others broke? Not virtue, and not luck. Beneath its politics it had a state that could hold. Since the reforms of King Chulalongkorn it had built one, a real civil service, an army that belonged to no single man, an idea of the nation older than any government on top of it. So when the violence of the age came inside, the structure beneath did not give: the taxes were collected the next morning, the currency held, the country could do the terrible thing it did and remain a country. Cambodia had rested its order on one prince, and when he fell there was nothing underneath to stop the Khmer Rouge. Laos had never been one country at all, only three factions under a name. A state is what lets a people survive its own history, and the one that had it paid in another coin.

Beneath even this lies the oldest constraint, which is not a matter of will. Thailand sits at the hinge of the mainland, between the Indian world and the Chinese, on the road any power that means to hold the region must travel, and it could not leave that place. The Americans who came in the 1960s thought in terms of a war and settled it inside a generation; the power rising now to the north thinks in terms of the map, and has lived beside that road for three thousand years. They believed they were holding a line that had to be held, and could not see, as perhaps no power outside a country can, the price the country would pay for being the ground it was drawn on. This is not a tale of villains but of consequence, of how the weight of an act outlasts the intention behind it.

I have watched a smaller version of this from the inside, and I have not been able to unsee it since. My working life was spent among very large powers, funds and states and the men who spoke for them, forever circling one another for advantage, and in that company I was the small party. I learned early what the small party learns. Lean too far toward one of the giants and another will remember it, and the price arrives later, quietly, in a form you did not predict. So I became careful. I gave little away in a room, kept my weight in the centre, made no move that could be read as choosing a side. It kept me standing in that world for a long time. What it cost arrived later: the man who can commit to no one belongs to nothing. The rooms and the names are not mine to give. But I know in the body what it is to survive by position rather than by strength. My own answer was to hold the centre; Thailand’s was to lean the whole way to one side. Both are answers the small give to the same danger, and both are paid for.

That is the thread that ties these three together, and it reaches far past them; a reader from the Middle East, or from anywhere a great power has fought across a smaller one, will know its shape without being told. A small country in a neighbourhood on fire cannot make the fire pass it by; that power it does not have. The one thing it can settle is the currency it will pay in. Thailand paid in alignment. It was not destroyed, and it is paying yet, inwardly, in a wound that does not show from the air.

Go back to the field at Thammasat, to the person holding the acrylic sheet up against the morning, lifting the old picture until its doorway finds the doorway still standing, the same stone, so that the dead lie for a moment over the living who cross the same grass to class. The image works only because the buildings never moved. The killing and the ordinary Tuesday share an address.

The runway is the same trick laid flat. Soon the jet from Kunming will not come down on a Cold War airfield; it will come down at an airport city, and the concrete poured to carry the heaviest bombers ever built will hold up a hotel. Nothing will be hidden. It will only stop lining up, the way a photograph stops lining up when you lower it from the light. The old man at the fence still knows where the seam is. When he is gone the strip will be a foundation and nothing else, and a foundation does not answer for what it was poured to bear.