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?

Read this before the season turns.

It was late when the shipping magnate leaned back from dinner — a private room in Abu Dhabi, overlooking the Corniche, the kind of place where the waiters know not to clear the coffee. His fleet moved more dry bulk tonnage than most countries consume. Over the last cup, he told me about the Bengal famine pricing — how in the 1770s, decisions made in small offices on Leadenhall Street had determined who could ship rice out of Bengal and at what cost, how the textile trade that had made Bengal the richest province in the Mughal world was unwound not by conquest alone but by the re-pricing of risk on Indian-owned vessels until local merchants could no longer compete on their own waterways. The looms of Dhaka fell silent not because the weavers forgot their craft. They fell silent because the financial architecture was withdrawn, one policy at a time.

Then a senior marine insurance broker — he had flown in from London for the meeting — said something I wrote down on the back of a briefing sheet and have carried since.

“Gentlemen, a navy closes a strait for days. An underwriter closes it for months.”

Nobody reacted. The room moved on to the next slide.

I did not.

Ibn al-Muqaffa understood this twelve centuries ago. In the Kalila wa Dimna — the ancient book of fables he translated from the Pahlavi, itself drawn from the Sanskrit Panchatantra — a merchant’s ship approaches a river crossing. He fears bandits and pirates. Hires armed guards. Watches the horizon. And while he watches, the river pilot — quiet man, knows the depth of the water — simply refuses to board. The cargo never moves. Not violence. The one man who made the voyage possible decided the terms were no longer acceptable.

I thought of that fable in late 2023, when Houthi drone strikes sent war-risk premiums spiralling and Maersk, MSC, and Hapag-Lloyd all suspended Red Sea transits within seventy-two hours. Not because the Red Sea was impassable. Because the insurance market had decided it was uninsurable. The missiles made the news. The premium adjustment made the decision.

• • •

This week, war-risk premiums spiked from 0.25 percent to over one percent of vessel value overnight — a fourfold increase. The P&I clubs — twelve mutual insurers underwriting roughly ninety percent of the world’s ocean-going tonnage — adjusted coverage. The strait, physically open, became economically impassable. Governments could theoretically self-insure — sovereign war-risk pools have been tried before, during the Falklands, during the tanker wars — but nothing at this scale, nothing covering fertiliser volumes for a billion farmers. Rerouting via the Cape of Good Hope adds fourteen days and roughly three hundred thousand dollars in fuel per voyage. For oil, that math works. For low-margin bulk fertiliser with a biological deadline, it does not.

Every great maritime power understood this. Venice withdrew the commenda from specific routes to destroy specific competitors — not as general policy but as targeted economic warfare. The East India Company created a two-tier insurance system: Company ships sailed fully covered while Indian-owned vessels faced rates so ruinous that the competitive position of Bengal’s merchant class was dismantled over decades. Lloyd’s codified the principle into empire: a trade route that could not be insured was a trade route that did not exist.

Venice, the East India Company, Lloyd’s — and now Hormuz. Insurance withdrawal is not a side effect of conflict. It is the primary instrument of economic control.

• • •

Now follow the chain the markets are not watching.One-third of globally traded fertiliser passes through Hormuz. India sources sixty-four percent of its urea imports from GCC suppliers — those imports doubled to nearly six million tonnes in the first half of last year. China imports roughly ten million tonnes of sulphur annually, fifty-six percent from the Middle East — feedstock for over half of China’s phosphoric acid production. When China suspended phosphate exports last year to secure its own food chain, the countries that had relied on Chinese phosphate as a backup discovered they had no backup at all.

Can Russia fill the gap? It cannot. Russia exports roughly forty-four million tonnes annually, but EU tariffs rising toward three hundred euros per tonne are redirecting Russian volumes into a contested scramble for Asian buyers. Black Sea ports face their own insurance premiums. Russia will gain share at the margins, but it cannot replace the volumes that transit Hormuz.

But the price is not the problem. The timing is. Nitrogen fertiliser is a physical input with a biological clock — the soil must receive it before the planting window opens. A shipment two weeks late does not produce a two-week delay. It produces an entire season of diminished yields. There is no strategic nitrogen reserve. Oil has the SPR. Fertiliser has nothing. The system was designed without a buffer because no one imagined the chokepoint would be closed by a spreadsheet.

Here is a number no one has published. If Hormuz closes during the pre-Kharif procurement window, India faces not merely a supply shortage but a fiscal detonation. In 2022–23, when urea prices tripled, India’s fertiliser subsidy bill hit ₹2.5 trillion. A prolonged Hormuz closure would push spot urea back to crisis levels — and the additional subsidy burden alone would add nearly a full percentage point to India’s fiscal deficit, pushing it into sovereign downgrade territory. For a five-acre farmer in Uttar Pradesh, the arithmetic is simpler and more brutal: the input cost increase from a single disrupted season can exceed the expected profit from his entire harvest.

• • •

Here is the cruel irony nobody in the commodity world discusses. Decades of synthetic nitrogen overuse have degraded the soil itself — depleted its organic carbon, killed the microbial communities that once made nutrients available naturally — so that the same field now requires more fertiliser each season to produce the same yield it once delivered with less. The soil becomes an addict, and the supply chain becomes its dealer. This is not a supply shock. It is a systems trap. The soil has been financialised into dependency on the very supply chain that just broke.

• • •

The transmission chain: fertiliser shortage, planting failure, grain spike, food insecurity, displacement, political instability, sovereign risk repricing. History is specific about what happens next. The 2007–08 food crisis, driven in part by fertiliser price spikes, triggered riots across thirty-seven countries — Haiti’s government fell, Egypt deployed the army to bake bread, Cameroon and Mozambique saw lethal street violence. In 2011, food prices helped ignite the Arab Spring — Tunisia, Libya, Egypt, Syria. Sub-Saharan Africa, where nations import the majority of their fertiliser and have the least fiscal capacity to absorb price shocks, remains the most exposed region on earth. The African Development Bank warned in 2022 that fertiliser shortages alone could cut continental food production by twenty percent. Allocators modelling Hormuz as an energy event are running the wrong scenario. This is a food-security event with an energy trigger. The timeline is not weeks. It is quarters.

• • •

The Mahabharata speaks of maya — the great illusion that what is visible is all there is. I was given all eighteen volumes by an old investor in Mumbai who had turned inward after forty years in markets. The ships are visible. The missiles are visible. But the actuarial calculation that closed the strait — made quietly, in an office that smells of carpet and old tea — is invisible.

Geopolitics loads the gun. The insurance market reads the trajectory. Underwriters are not the cause — they are the signal. The most transparent, most honest transmission mechanism the global economy has. But it is their signal, not the diplomat’s or the admiral’s, that determines whether ships sail.

The Venetians knew. The East India Company built an empire on it. Those who control the terms of trade control the fate of nations.

The question is why the rest of the world has forgotten that the most powerful weapon in maritime trade has never been the warship.

It has always been the ledger.

• • •

In a world where a chokepoint can be closed by an actuarial calculation rather than a military command, who underwrites the food supply of two billion people? And what happens when they decide the risk is no longer worth carrying?

I have looked for that question in allocation models. I have not found it.

That silence is the risk.

— Nazem

Commentary only — not investment advice. Author has no position in referenced instruments. Past data does not predict future performance.

Nazem Alkudsi, CFA, is the founder of @LongArcNews. A former CEO in the Abu Dhabi sovereign wealth ecosystem and four-decade veteran of institutional investing, he writes about capital, power, and civilisational patterns.

The mine was closing and the geologist was laughing.

Not bitterly. Not yet. It was a social gathering in Park City, Utah—late January, the kind of weekend where skiing is the excuse but the conversations are the real event. Overcast sky. The town smelling of woodsmoke and cold pine. I found myself in a long exchange with the head of the Department of Geology and Geological Engineering at the Colorado School of Mines—the oldest earth sciences program in the country, founded in 1874 in Golden, Colorado, back when what lay inside the mountain was the only thing that mattered.

He told me something I have carried ever since.

At the height of the dot-com fever—right at the millennium’s turn, just before the whole thing came apart—his school had cratered to its lowest enrollment in living memory. Faculty leaving. Students flooding into computer science, finance, anything that didn’t require thinking about rock. The world had made up its mind. The future was digital. Virtual. Weightless. The earth was merely a platform. Not a partner.

The numbers told the story plainly. Mines enrolled roughly 3,787 students by 2002—a trough before a decade of slow, grinding recovery. Nationally, geoscience enrollment tracked the price of oil, and oil was cheap, and software was king. Seven percent of American high schoolers were taking an earth science class. Geology departments shuttering across the country. When students were surveyed about career paths, geology ranked dead last.

That conversation did something to me. I was a sovereign fund professional—capital allocation was my world, not mineral extraction. But I couldn’t shake what he’d said. So that year, knowing almost nothing about the subject, I got interested in precious metals mining. Arranged visits to a handful of silver mines in northern Idaho—the Coeur d’Alene mining district, the Silver Valley, where fortunes had once been pulled from the Bitterroot Mountains and whole towns had grown up around the shafts.

What I found was not a sector in decline.

It was a world being erased.

One mine shutting down. Another weeks from the same fate. The towns that had grown up around them—places like Kellogg and Wallace, towns that existed for no reason except that the mountain gave them one—going bankrupt. Main streets half-shuttered. You could see it on the faces in the diners, in the hardware stores that had nothing left to supply. Not anger. Something quieter. The look of someone watching the only life they’ve ever known fold up and disappear, knowing nobody outside their valley has noticed. Knowing nobody particularly cares.

Silver was trading below five dollars an ounce. The world treated it as a relic—a monetary metal from another century, of interest to coin collectors and no one else. Nobody thinking about silver as an industrial metal. Nobody asking what the most electrically conductive element on the periodic table might be worth in a future built on circuits, sensors, and servers.

I remember standing in one of those towns on a grey afternoon, looking up at the mountains, thinking: the world has decided that what comes out of the ground no longer matters. That the future belongs to code, to networks, to things you cannot hold in your hand.

The world had stopped looking at the ground.

Twenty-four years later, those memories return with force. The conversation in Park City. The silence inside the closing mine. The faces in the Idaho towns. Because every AI lab being built in the Gulf, every data center humming in Virginia, every sovereign fund pouring capital into the digital future—all of it depends on three things no algorithm can conjure: rare earth minerals locked inside mountains, fresh water to cool the machines, and copper wire to carry the current.

Rock, water, and wire.

The school in Golden now enrolls over 8,200 students. Some of those Idaho mines have reopened, their silver and cobalt suddenly strategic again. Silver? It trades near eighty dollars an ounce today—nearly twenty times what it fetched when the world decided it didn’t matter.

The world remembered. But only after it had to.

Everyone frames the AI race in software terms: US versus China, open-source versus closed, Nvidia versus AMD. Model architecture. Parameter counts. Benchmark scores. These are the metrics that fill the front pages and the pitch decks. Meta, Amazon, Alphabet, and Microsoft have collectively committed to spend nearly $400 billion in 2025 alone. BlackRock and Microsoft are assembling a $100 billion AI infrastructure partnership. Three sovereign wealth funds just co-signed what may be the largest AI infrastructure check in history—but the real story is not the headline number. It is the $50 billion in physical infrastructure commitments and the electricity guarantees that came with it.

The capital is extraordinary.

But these are software framings backed by financial engineering. And the long-arc view says the real competition is not financial at all.

It is geological.

The question is not who builds the smartest model. The question is who controls what the model is built on.

Rock

The Vedic tradition treats the Earth herself—Prithvi—as a goddess whose body bears the weight of civilization. The Atharva Veda’s Hymn to the Earth calls her “mother” and asks that what is drawn from her be quickly replenished—a prayer that sounds less like devotion and more like a term sheet between humanity and the ground it stands on.

The global critical minerals market hit $328 billion in 2024. Projected to reach $586 billion by 2032. The International Energy Agency says meeting this rising demand will require $500 billion in new mining investment by 2040. Copper faces an implied supply deficit of thirty percent by 2035. Lithium, forty percent. These are not figures from activists or venture capitalists talking their book. They are from the IEA—an organization not known for alarmism.

And then there is silver—the metal the world forgot and is now scrambling to remember.

Silver possesses the highest electrical conductivity of any element and the highest thermal conductivity of any metal. Irreplaceable in the AI buildout: threaded through the high-performance chips, the GPU packaging, the thermal interface materials that keep processors from burning out, the silver-plated connectors and switchgear that distribute power through data centers, the solder that bonds semiconductor to server. A single 500-megawatt solar array—enough to power one hyperscale data center—requires roughly 300 metric tons of silver.

Oxford Economics, in a report commissioned by the Silver Institute, recently designated silver a “next-generation metal,” essential across every high-growth sector from AI to electric vehicles to clean energy. The silver market has been in structural supply deficit for five consecutive years, with the world consuming 160 to 200 million ounces more each year than the mines can pull from the ground. At five dollars an ounce in 2002, the market was telling you silver was a relic. At eighty dollars today, after a 148 percent gain last year alone, it is telling you something very different.

Any serious portfolio positioned for the physical infrastructure age and still without meaningful silver exposure is, in my view, incomplete.

The geography of control has barely shifted. China still refines over eighty percent of the world’s battery-grade graphite and rare earth elements—not because it has the most deposits, but because it built the processing infrastructure during the decades when the rest of the world was busy building software platforms and pretending the periodic table was someone else’s problem. The United States just took a ten percent equity stake in USA Rare Earth as part of a $1.6 billion deal. Canada launched a C$2 billion Critical Minerals Sovereign Fund. The Trump administration is exploring a dedicated sovereign wealth fund for domestic mining.

Call them trade policies if you like. They are acts of geological sovereignty, undertaken with the quiet urgency of governments that have realized, very late, that they outsourced something they cannot afford to have outsourced.

The pace of this reckoning is accelerating faster than most allocators grasp. Fifty-four countries convened for the first-ever Critical Minerals Ministerial—a gathering that, had it occurred in the oil markets, would have been front-page news for a month. The PIF—Saudi Arabia’s Public Investment Fund—has begun pivoting toward minerals, signaling that the Gulf’s sovereign capital is no longer content to finance the digital layer alone. It wants the geological layer beneath it.

What we are witnessing is not supply-chain diversification. It is the emergence of a parallel pricing system. Friend-shored minerals will trade at a premium. Those outside the trusted network face stranded assets and shrinking markets. This is the Bretton Woods of geology—and the sovereign funds that see it early will co-invest on terms that did not exist six months ago.

Here is the part that should trouble us most: mining investment rose just five percent in 2024, down from fourteen percent the prior year; adjusted for cost inflation, real growth was a mere two percent. The capital pouring into AI models dwarfs the capital going into the ground those models depend on by an order of magnitude.

This is not a market inefficiency. It is a civilizational blind spot—and not merely an intellectual one. A moral one. We are building the future on a foundation we have collectively decided is not worth investing in, while the communities that live above that foundation—the miners and the engineers and the geologists and the small towns in Idaho and Montana and the Congo—bear the consequences of our negligence.

The sovereign fund that secures offtake agreements on critical mineral refining capacity today holds the equivalent of what upstream oil concessions were in the 1970s.

The allocator who sees this is not early. They are merely not late.

Water

The Quran says: “Do you not see that Allah sends down rain from the sky and makes it flow as springs in the earth?” The image is not metaphor. It is hydrology—and theology—in a single sentence: water is given, not generated. And the giving has conditions.

Jordan just committed $6 billion to a single desalination plant. Forty percent of a nation’s drinking water, from one facility.

Sit with that.

Think about what it means for a family in Amman when they turn on the tap—the knowledge, conscious or not, that their water comes from a single point of engineered vulnerability, that the margin between normal life and crisis is exactly as wide as one plant’s operational continuity. This is what sovereignty looks like when water is scarce. It looks like a single point of failure dressed up as infrastructure.

A large AI data center consumes up to five million gallons of water per day—equivalent to the daily needs of a town of 50,000 people. Google alone used over six billion gallons across its data centers in 2023, with nearly a third drawn from watersheds already classified as stressed. AI workloads consume ten to fifty times more cooling water than traditional servers. In Texas alone, data centers are forecast to consume 399 billion gallons annually by 2030—up from 49 billion this year. The United Nations recently declared the world heading toward “global water bankruptcy.”

Institutional finance has no water beta. No water-adjusted discount rate. We have priced carbon. We have priced volatility. We have priced credit risk down to the basis point. But the water that makes the digital revolution physically possible—the one resource without which the cooling systems fail and the chips overheat and the data centers go dark—has no price.

The spreadsheet has no line for it.

Think about what we are saying when we leave that line blank. We are saying that the thing without which none of this works is worth, for modeling purposes, nothing. Some will point to closed-loop cooling systems and corporate pledges—Microsoft, for instance, has committed to replenishing more water than it consumes. But such commitments remain the exception, not the standard, and even recycled water creates concentrated thermal and hydrological stress on local watersheds that dispersed usage does not. The problem is not solved by recirculating it. It is relocated.

The Buddha asked: “Do you have the patience to wait until your mud settles and the water is clear?” The question was about the mind. But it might as well be about the markets—and the resource they have refused to see.

Wire

The hottest debate among institutional investors right now—the one happening not in published research but in group chats and allocation committees—is about voltage transformers. Lead times of two and a half to three years. Grid expansion cannot keep pace with data center demand.

The scale of the mismatch is breathtaking.

The big five tech companies are expected to spend over $600 billion in capital expenditure in 2026 alone. Data center construction has surged fivefold in two years, now surpassing spending on all other commercial buildings combined. Global data center electricity consumption jumped from roughly 240 TWh in 2022 to 415 TWh in 2024—nearly doubling in two years. US data centers today consume 75.8 gigawatts. By 2030, that figure reaches 134.4 GW. Goldman Sachs projects data center power demand will grow 175 percent by the end of this decade. Every one of those figures describes commitments already made, contracts already signed, ground already broken.

And what carries all that power? Copper. A single hyperscale AI data center can require up to 50,000 tons of it. Nvidia’s DGX GB200 NVL72 rack alone contains over two miles of copper NVLink cables. Yet chips are being warehoused across the country because there is no power to run them. Grid connection lead times run one to two years.

The AI buildout is running into a wall that no amount of capital can scale: the electrical grid itself.

The hyperscalers know this. They are not waiting for the grid to catch up—they are building around it. Meta is betting on nuclear through partnerships with Oklo and TerraPower. Data centers are deploying repurposed aircraft engines as on-site power. Bloom Energy fuel cells proliferating across campuses that cannot secure grid connections fast enough. Strip the press releases away and what you see is an admission, made with billions of dollars, that the existing power infrastructure cannot support the future these companies have already sold to shareholders.

And it creates a second-order competition: not just who has grid access, but who can generate power independently of the grid entirely.

There is a constraint that no amount of engineering can bypass: political will. Grid reliability costs have spiked. Virginia is now forcing data centers to pay eighty-five percent of transmission infrastructure costs. Moratoriums on new data center construction are emerging in communities from northwest Indiana to rural Virginia, backed by coalitions that cross every political line—united in the simple proposition that ordinary families should not subsidize the electrical appetite of trillion-dollar companies.

The geology competition may ultimately be won not just by the nation with the most rock, the most water, and the most wire—but by the nation with the political will to build.

Here is the detail that should keep allocators awake: the corridors where sovereign-backed data centers are being built fastest—Texas, Virginia, Arizona—are the same corridors where the ground is sinking. Aquifer depletion, land subsidence, and grid congestion are converging in the exact geographies where the AI buildout is most concentrated. We are building the most advanced computational infrastructure in human history on land that is physically subsiding beneath it.

The infrastructure is being priced on assumptions the geology no longer supports.

Genesis records the first mandate given to humanity: to work the earth and take care of it. The instruction was not to extract or exhaust, but to tend—a word that implies obligation, not ownership. The Psalmist added: “The earth is the Lord’s, and everything in it.” We are tenants. And tenants who destroy the property eventually lose the lease.

The irony is almost too perfect: the most advanced technology humanity has ever built depends entirely on the most basic infrastructure humanity has ever struggled with. Consider the convergence now underway—the United States becoming Intel’s largest shareholder in the same season that fifty-four countries convene to form a critical-minerals trading bloc. These are not separate stories. Semiconductor sovereignty and mineral sovereignty are the same story—the physical layer asserting itself against a world that had been pretending the physical layer was optional.

Every tradition understood what the spreadsheet does not: the physical world is not an input to be optimized. It is a partner to be respected.

And partners have terms.

We forgot this once—in Golden and in Idaho, where the mines went quiet and the spreadsheets had no line for what was lost.

The ground did not forget.

What Could Prove This Wrong

Any honest thesis must name the forces that could undo it.

Technology substitution is real: the solar industry is reducing silver loading per cell, copper recycling is improving, and small modular reactors could begin to relieve the grid bottleneck within the decade. An AI winter would weaken the timeline for every projection cited here. These are possibilities any serious allocator must weigh.

But here is why I believe the structural case survives them.

Substitution reduces intensity per unit—but the number of units is growing exponentially. Solar installations are multiplying faster than silver loading is declining; data centers proliferating faster than cooling efficiency is improving. Recycling helps at the margin but cannot close a deficit measured in hundreds of millions of ounces when above-ground inventories are already depleting. As for nuclear: even the most optimistic SMR timelines put meaningful grid contribution in the early 2030s, and the permitting, construction, and interconnection challenges are formidable.

The grid constraint is not a five-year problem waiting for a five-year solution. It is a five-year problem facing a ten-year solution.

An AI winter, while possible, would not eliminate the electrification megatrend, the clean energy transition, or the defense and industrial applications that drive mineral demand independently of AI. The AI buildout is the accelerant. Not the sole engine.

What the bear case actually reveals, when you press on it, is the same thing the bull case says: the physical layer is the binding constraint. Whether demand grows at the pace the optimists expect or at half that pace, the supply side is structurally unprepared. Mining projects take seven to fifteen years from discovery to production. Refining capacity takes nearly as long. These are not timelines that can be compressed by enthusiasm or capital alone.

The bear case slows the clock. It does not change the physics.

One more counterargument worth engaging, because it is the most seductive: that AI itself will solve the geology problem. Companies like Earth AI are already using machine learning to discover mineral deposits at a fraction of traditional cost—$2 million per site versus $100 million through conventional exploration. The technology is real. It is impressive. But finding is not having. Faster discovery does not solve the refining bottleneck, the permitting timeline, or the seven-to-fifteen-year development cycle. AI can find the deposit. It cannot will the mine, the smelter, or the political consent into existence.

The very technology accelerating demand for minerals is also accelerating the discovery of how much we lack—without accelerating the capacity to produce them.

This is the paradox of acceleration: the faster AI advances, the more clearly it illuminates the constraints it cannot overcome.

The allocators who built positions in upstream oil in the early 1970s, before the embargo, before the repricing, before the consensus caught up—they did not do so because they predicted a specific catalyst. They did so because they understood the structural mismatch between demand and supply, and they understood that such mismatches do not resolve gently.

The trigger is unknowable. The direction is not.

The consensus says AI winners will be determined by model architecture, capital, and talent.

The Long Arc view: AI winners will be determined by who controls the rock, who secures the water, and who owns the wire. The software layer is a distraction. The real allocation question for the next decade is physical—and most institutions are still underweight it. A $586 billion critical minerals market by 2032. A thirty percent copper supply deficit by 2035. A silver market in structural deficit for five straight years with no substitute for its conductivity. Data center power demand growing 175 percent by 2030. Nearly half of all data center facilities facing high water stress by mid-century.

These are not tail risks. They are the base case.

I wonder if you’ve noticed what I’ve noticed. That the conversations about AI have become almost entirely abstract—models, parameters, benchmarks, valuations—while the physical reality underneath grows more fragile, more contested, and more consequential by the month.

The Gulf states understand this better than most. They sit at the intersection of all three constraints—investing billions in AI infrastructure while managing acute water scarcity and building desalination capacity that doubles as strategic insurance. When the UAE launched its sovereign AI model in January 2026, it was not merely a technology announcement. It was the declaration of a chain of custody that begins in the mine and ends in the model.

The Gulf states have turned desalination into a civilizational achievement—making life, agriculture, and industry possible in one of the most water-scarce regions on earth. But that achievement carries a cost the region’s own scientists are increasingly confronting. The Arabian Gulf is now roughly twenty-five percent saltier than typical seawater, with hotspots running double or triple normal salinity. The GCC nations produce fifty-five percent of the world’s desalination brine—142 million cubic meters of hypersaline concentrate discharged daily into coastal waters, with the Gulf receiving the largest share. The ecological effects are measurable: elevated seafloor salinity, stress on coral systems, and an energy footprint that makes desalination one of the region’s largest sources of emissions.

The Gulf governments know this. Several are investing in next-generation brine management, zero-liquid-discharge systems, and renewable-powered plants. The nations that solve the brine equation will hold a strategic advantage as consequential as the nations that control the minerals—and the Gulf, precisely because it depends on desalination more than anyone, has more incentive to lead that transition.

Water sovereignty and marine stewardship are not competing priorities. They are, for the Gulf, the same priority.

The Buddhist tradition teaches pratityasamutpada—dependent origination—the principle that nothing arises independently, that every phenomenon exists only in relation to the conditions that produce it. The most advanced neural network is no exception. It depends on the mountain, the aquifer, and the copper seam.

Sever the conditions, and the phenomenon ceases.

This is not philosophy. It is physics.

The future is not being built in the cloud. It is being pulled from the earth, carried by the wire, and cooled by the water.

And the earth, as always, will have the final word.

If this resonated, read next: The Friday Deadline