As of Thursday the 27th, Brent is about $86, having shrugged off a PCE report that came in around where analysts expected. And in about two hours Kevin Warsh gives his first Jackson Hole keynote as Fed chair, into a market that has spent the week arguing with itself about whether the next move is a hike, a hold, or a cut. (finishing this the night before and printing it to the 8 AM est release slot)

The recession nobody stumbled into

The 1982 recession gets taught as a price the United States, or at least, its political and economic elites paid to break inflation… (not that they were the ones who shouldered the burden) But less attention is paid to the role the Volcker shock had on the American global economic order. Granted, the early 1980s is not like the current headwinds facing the American order in the mid-2020s. But this bit of history should be indicative of the past as a guide to the future.

The United States is Picking up its ball and going home

as mentioned in the earlier series, Oil going ‘higher for longer’ does not land evenly. It lands hardest on developing economies that import the barrel and borrow in dollars to pay for it, and it lands nearly as hard on the American ‘junior partners’ — the EU, Japan, South Korea — whose entire industrial model assumed cheap imported energy and open sea lanes. Then add the supply chains already cut into: fertilizer and refined products, hit either by the Iran war or by the strike campaign American-backed Ukrainian proxy forces are running against Russian refining, grains, et al.

Take them together and it stops reading as a series of accidents. It is a multipronged attack, and its function is to end a specific set of arrangements — the form of globalization Washington built and imposed on the world from the 1990s through the 2020s. America is picking up its ball and going home.

Two machines, one rate shock

The domestic history is straightforward. The 1970s inflationary spike was defeated by a high rate regime, brought on by a newly independent federal reserve, chaired by Paul Volcker, who was installed in 1979 (Side illustration for the bipartisan consensus here). The U.S. domestic economy imposed austerity and a recession to discipline the now nostalgically referenced American working class, particularly unionized factory workers. But that is just the domestic side of the story.

There were two machines and one shock, and they are worth keeping separate. In the foreign policy domain, the debt crisis worked on what the periphery owed — dollar obligations repricing upward against borrowers who earned in something else. The commodity bust worked on what the periphery sold — export revenue repricing down at the same moment the debt service repriced up. For many developing economies in the Global South, being caught in both is not simply a downturn. It is a vise.

The pain at home was real. The pain abroad was worse, and the gap between them was the product.

The United States absorbed a hard recession and came out the other side with disinflation, cheap commodity inputs, and financial dominance reasserted — then spent the back half of the 1980s buying a discounted world. A rate shock is just one more monetary tool at home and another instrument of hegemony abroad, and the second function is what I am musing on today, in a period where so much financial chatter concerns the bond market, the U.S. national debt, and of course the proverbial dollar debasement trade — which presumes the dollar is destined for collapse and decline as the global reserve currency, and will drive all commodity prices through the roof.

I am of the mind that what is coming will be similar, though not exact. (I do not anticipate interest rates at 15 percent). But the dirty Machiavellian reality ahead is the same one: discipline the periphery, cheapen its output, then harvest it.

Why the distant history fits better than recent history

Much of the so called ‘debasement’ trade presumes a U.S. recession will be met with a similar script to the global financial crisis of 2007–2008 — namely quantitative easing, and the metaphor of printing money, thus driving the value of the dollar down against everything else, equities and commodities in particular.

But that took place under completely different economic conditions, with U.S. policy planners choosing to reflate the existing sectoral leaders of the global economy.

The ‘BlitzCorrection’, as I've conceptualized the idea, plays into my counter thesis. It is a deliberate, (or at the very least, an acceptable pain of doing business) demolition — a managed rotation of capital out of the factions of the rentier capitalist coalition and into the physical-industrial one. I’ve been open to discuss, and self reflect on whether or not this rotation can occur without the blitzcorrection. I have higher conviction on where we’re headed than how we get there, but for now, I still see no way to rotate without deflating certain asset classes. Whether we get there slow or fast, that rotation needs two things a Volcker-style shock supplies natively:

Cheap inputs. An industrial resurgence is an input-cost problem before it is anything else. A global deflationary bust delivers cheap energy — which the U.S. now domestically leverages as an energy superpower — cheap metals, and cheap capital goods. Those are the conditions under which rebuilding a domestic manufacturing base pencils out. A reshoring program wants the world’s commodities on sale, and a rate shock alongside an induced global recession is one of the ways it gets there.

Duration selectivity. Semiconductors and AI-growth equities are the longest-duration assets on the board — most of what they are worth is money nobody expects to arrive for a decade. A rate shock discounts future cash flows hardest exactly where those cash flows sit furthest out, because the discount compounds across every year in between. Which means the same instrument that disciplines the periphery also craters the speculative multiple. Or to put it in my more political-economy and factional terms: If the so-called free market will not subordinate or promote the chosen sectors of capital, then a broad industrial and financial policy will be marshalled in to help the market see the light.

To be clear, in case there is still any confusion and the two mechanisms get confused: a rate shock is the amplifier here, not the break. A purely financial shock gets absorbed, and two things do the absorbing — what is called dealer gamma, and the passive bid that Mike Green has written and spoken on at length. Between them they have already absorbed several.

Dealers hedge the options they sell by trading the underlying asset, and positioned the usual way that means buying dips and selling rallies — a mechanical dampener. Short gamma flips it, and the same hedging sells into weakness instead, which is why the absorption holds right up until it doesn’t.

What none of that machinery can do is absorb a physical chokepoint. That is what cannot be routed around, and the rate shock’s job is stripping the valuation cushion sitting on top of it.

Our dollar, your problem

There is one claim in the Brent Johnson dollar-milkshake world worth taking, and it is worth taking exactly that far — though the mechanism under it is considerably older than the metaphor. Hélène Rey took it to Jackson Hole in 2013, arguing that American monetary conditions drive a global financial cycle no matter what exchange-rate regime anyone else runs, and the BIS has spent years documenting the channel underneath it: when the dollar strengthens, cross-border lending contracts and investment in the borrowing economies goes with it. The claim: a rising dollar is a reading on the rest of the world's condition, taken with an American thermometer. Offshore borrowers who owe dollars and earn something else have to scramble to find dollars; the scramble bids the currency up; the climb registers distress that is happening somewhere other than the United States.

The lay reading runs the other way, treating dollar strength as a report card on American health. That inversion is the bridge from the monetary mechanism to the geopolitical program, and everything downstream depends on getting it the right way round.

A strong dollar is a fever. The fever belongs to somebody else. Or to trot out more economic clichés of old: when the U.S. economy catches a cold, the rest of the global economy catches a fever. But unlike an accidental cold, this one looks like it is being self-imposed from within. A strong dollar is a fever, and the thermometer is American, and it is up almost everybody else's ass.

The modern inversion — a spike, not a plateau

As noted above, I do not anticipate an exact replay of the Volcker shock. I am looking at the tools in the toolshed.

The 1982 analogy carries one difference that reshapes how the move could be run at all. Volcker could hold rates high for years because the sovereign could take it: federal debt sat at roughly a third of GDP. It is north of 120 percent now. The United States is the most rate-sensitive borrower in the room, and a sustained plateau would detonate the Treasury’s own interest bill well before it disciplined anyone abroad.

So a modern rendition cannot be a siege. It would have to arrive as a demolition charge — a spike, then break the layer, then relent. The fiscal arithmetic forbids duration, which buys Volcker’s violence without Volcker’s patience: a sharp shock timed to the cascade, then a fast retreat once the speculative layer has broken and the rotation is underway. The rate regime can be reset afterward to whatever the recovery phase requires. This is a tool for a moment of maximum pressure rather than a policy stance to be held.

In saying this, I am thinking more about the coming sequence of events. Dollar strength as a means to an end, not an end in and of itself. Once the damage is done, and the Treasury and the Fed and the broader economic planners are content, lowering the dollar exchange rate may very well — and in fact will likely — be reimposed.

Two features of the current setup fit that reading. Forward guidance has been removed, which is the precondition for a surprise. A central bank reserving the right to shock stops telegraphing first.

The second takes a moment to spell out. The Fed has two levers, not one. The first is the short-term interest rate, the number everybody watches, and it is also the number that sets what the Treasury pays on its own new borrowing. The second is the balance sheet: whether the Fed is buying bonds and putting money into the system, or letting them mature and draining it back out. Warsh’s preference runs to the second lever.

That distinction is the whole trick. Draining liquidity pushes up long-term borrowing costs — mortgages, corporate debt, and the rate against which every future dollar of profit gets discounted — without the Fed having to lift the short rate to a level that detonates the government’s own interest bill. The damage lands where the demolition wants it, on long-duration assets and speculative multiples, and it skips the one place the sovereign cannot afford to be hit. The demolition gets delivered through the liquidity channel, with a short rate jolt on top.

The excuse

In the age of Trump, pretexts and the information channel are flooded with noise and contradictory statements masking intent that is driven not simply by the administration, but by the bipartisan consensus resting underneath American policy making.

Despite all this, a deliberate hike needs cover. No central banker announces a demolition, and a hike delivered into a calm tape reads as either incompetence or something worse.

An inflationary energy shock supplies that cover. A hike delivered into an oil pinch reads as orthodox inflation-fighting — the most conventional move in the book, made for the most conventional reason — and orthodoxy is the only register in which the move is available at all. The excuse is not decoration on the mechanism. It is a precondition of it.

The energy shock does double duty: it pays the thesis, and it writes the permission slip for what I have been framing as the permission layer.

And the pressure is not hypothetical. The Fed held at 3.50 to 3.75 percent in July, but three policymakers dissented in favor of a hike — the most dissents in that direction since September 2016 — from Cleveland’s Beth Hammack, Minneapolis’ Neel Kashkari, and Dallas’ Lorie Logan. The June projections carried nine officials expecting at least one hike this year, with the median path revised up to 3.8 percent by year end and the inflation projections revised up alongside it.

September hike odds have since fallen from around two-thirds to roughly a third, and how they fell is the part worth sitting with. About half the decline was in by the sixth of August, on Hormuz de-escalation headlines — before either of the soft American data prints that got the credit afterward. The rate path repriced off the oil channel first and the payroll number second. That is the permission structure showing up in the tape rather than in my argument.

Which is to say the possibility is fully priced and openly argued. Some of the desks are now talking about a cut. That is a live debate and I am not pretending to settle it here. What nobody is arguing about is what a hike would be for.

The floor under the dollar

The obvious objection: a temporary shock followed by a relent should hand the dollar back everything it gained. It will not, because the dollar carries structural bids underneath it that survive the retreat.

• The Natty Greenback layer. A closed sovereign LNG loop — Henry Hub pricing, long-term supply contracts, an export-authorization gate that runs by adjudication rather than by rule — manufactures dollar demand that does not care what the rate cycle is doing. This is not the petrodollar, and it will not be for years; the volumes are nowhere close. What it has instead is duration. Those contracts run into the 2040s, which is a different kind of support than a flow that can reprice in a quarter.

• The digital-dollar layer. Dollar-denominated stablecoin rails extend dollar demand into channels the 1982 dollar never had, deepening the base of holders who need the currency for reasons unrelated to its yield.

• Capital-market gravity in the AI race. The deepest capital markets and the clearest path to buildout financing both sit in the United States, and capital chasing that advantage has to transit the dollar to get there. Being the venue is itself a bid.

Layer those under an acute funding-scramble spike and the sequence will not run spike-then-collapse. It will run the spike, break the layer, and then relent — onto a higher floor than where it started. The shock would be the exit signal for a long-dollar trade; the structural bids are why the dollar as a system will not break on the far side.

And as a point of reference, a weak dollar exchange rate is not a sign of decline in dollar hegemony to begin with. It is a sign of what American global financial power wants from its clients and trading partners at a given moment. In a few years, with a U.S. export regime up and running, a weaker currency will help that export regime in the same way the tool has been used historically by Japan or China. Weak exchange rates help pricing power in foreign markets.

The selectivity is the point

The Volcker shock victims were dollar-indebted emerging markets. China is not one of those — sovereign-solvent, reserve-rich, able to insulate. So a modern rate shock would discipline the allies and the vassals far harder than it disciplines the rival. Europe, Japan, the dollar-borrowing states of the Global South: the coalition Washington already controls is the coalition that gets milked, and the rival remains the harder target it has been in every other chapter of this.

And of course, as previously noted on this page, it will also create considerable economic stress on China’s trading partners, since China has built up a highly successful geopolitical economic fortress internally. Its export markets, on the other hand, are not so immune.

A basis point is one hundredth of a percentage point — the unit central bankers argue in, and twenty-five of them make the quarter-point move everybody is waiting on.

To repeat my Kissinger phrasing in basis points: To be an ally is fatal.

Honest friction—anticipated critiques

In 1982 the world had nowhere to run but dollars. Weaponizing the dollar now — the sanctions, the blockade, the coercion — accelerates the search for alternatives, which means a dollar shock has diminishing returns and could hasten the de-dollarization that breaks the dollar on the far side of the crisis. This is what is often cited when it comes to the lessons learned by Russia after the Ukraine invasion. That is the open question. This question is major support for the dollar debasement crowd, and the logic is admittedly seductive, so I get it. But I obviously think this thing is going the other way.

Which brings me to the part of this I think the ‘goldbugs’ have backwards. There simply is not enough depth in the gold market to efficiently supplant the dollar without foreign central banks being forced to wildly overpay in a market they would be driving higher themselves. So for all the foreign-central-banks-are-buying-gold narratives floating around, I suspect they are buying because price levels have made it a sensible investment — not because they have found an infinitely long-term solution. That is a crowded trade in a market too small to hold what would be asked of it.

The bottom line

The Volcker option is available and the setup rhymes: A fed chair inflation hawk with no forward guidance, a balance sheet he would rather drain than refill, an inflationary energy shock already in motion, and a reshoring program that wants the world's commodities cheap. The debt load forbids a 1982-style plateau, so if it comes it will come as a demolition charge — spike the interest rate, break the speculative layer, relent — and the structural bids under the dollar are why it lands on a floor.

This should not be read as a scheduled event, and I wouldn’t expect it today. It is simply one more instrument the U.S. architectural planners can reach for to break the rentier layer hard and cheap inside a single window.

The market has now priced the possibility of the move. What it has not priced is the reason.

Editor’s note: I use Claude Code for data collection, analysis, mathematical modeling, drafting, and formatting (in varying degrees and times). High-level architecture is my own.

Third in a series of three. The first piece laid out the map: three machines on three clocks, and one structure underneath them. The second took the gas. This one takes the oil, and it assumes the frame rather than rebuilding it.

As of this writing, the market is in one of its peace-is-at-hand phases, and Saturday is the reason to distrust it. Brent broke below $80 on 4 August, the first time since the war reignited, and closed the week near $83, with West Texas Intermediate at $78 — down about 7% on the week and up around 6% on the month, which is the whole problem stated in one line. Hold on to that gap. In March it averaged $11 and touched $25.

Trump said on 5 August that the Strait of Hormuz would reopen soon or Iran would be struck again. Then, inside a single Saturday: Tehran announced the framework of an Oman agreement was finalized, Iran’s Supreme National Security Council published six conditions and said the strait stays shut until American behavior changes, and a tanker belonging to Abu Dhabi’s national oil company was hit inside the strait.

I have called this Schrödinger’s ceasefire — an arrangement held in both states at once because both states are useful. What Saturday adds is that somebody did declare it over, in six numbered points, and it did not count. The machines that price crude do not hear the Iranian side: Washington plays with text, cheaply and as often as it likes, and Tehran can only answer with action. One thing was different this time. The New York Times carried the demands by Saturday morning. That could be noise. It could also be the first crack in the asymmetry.

This phase may hold. Maybe it lasts two weeks, maybe it lasts a month, or through the midterms. I have my doubts it goes that long. But nothing below turns on which, and that is deliberate — the argument runs on a sustained price, and a sustained price is not a headline.

Not the lights

Start by killing the version of this argument that gets it wrong, because it is the version most people reach for.

Expensive oil does not dim anybody’s grid. Oil generates about 2.5% of world electricity — a band of roughly 2.4% to 2.9% depending on whose aggregation you use — and the share has been falling for decades. Oil-fired generation ran near 1,324 terawatt-hours in 2000 and roughly 786 by 2023. Say “expensive oil makes their power expensive” in a room with an energy analyst or geostrategic planner and you lose the room in one sentence.

The exceptions have names, and for mostly predictable reasons. But even these are on a ticking clock of future reduction and green transition. The Middle East burns around 1.5 million barrels a day for electricity — roughly 40% of the world total, mostly Saudi and Iraqi summer air-conditioning peaks — and Saudi Arabia’s Liquid Fuel Displacement Program is scheduled to remove about 1 million barrels a day of that by 2030 in favor of Jafurah gas and renewables (IEA, Oil 2024, pp. 27–28). The one place on earth where crude still meets the grid at scale is inside the producer bloc, and it is scheduled to exit the practice this decade. The other exception is island grids — oil at roughly 72% of generation in Cyprus, 61% in Hawaii — which are economically brutal for the people living on them and macroeconomically invisible.

Here is the useful way to hold it. The 1970s shocks did their work. Grids de-oiled, and that substitution is complete, which is why 2026 is not 1973 for the lights. But the truck fleet, the tractor, the steam cracker and the airliner did not de-oil. They still run on the barrel, and there is no ready substitute for any of them inside a decade.

So the friction is real. It’s just not in the electricity layer. And to repeat my grand strategy thesis: the friction is the point.

What the barrel actually is

A barrel of crude is not a fuel. It is a slate of products, and each cut lands on a different part of an economy. The 2024 breakdown, out of 103.2 million barrels a day of world demand (IEA, Oil 2024, demand table, p. 14):

• Diesel and gasoil, 27.5% — trucking, rail freight, farming, construction, mining, marine gasoil, backup generators

• Gasoline, 26.4% — passenger cars and two-wheelers

• LPG and ethane, 14.5% — chemical feedstock, plus cooking fuel for 2.3 billion people in Asia

• Jet fuel, 7.3% — aviation

• Naphtha, 7.3% — steam-cracker feedstock: plastics, fibres, aromatics

• Residual fuel oil, 6.3% — ship bunkers

• Everything else, 10.7% — asphalt, lubricants, petroleum coke, waxes

Road transport is about 45% of the barrel. Industry is about 20%, two-thirds of that as chemical feedstock. Power generation is under 4% (IEA, Sheltering From Oil Shocks, March 2026, pp. 8–17).

From that slate, four channels carry a crude shock into an economy. In order of macroeconomic weight:

The import bill. This is the fastest and the broadest, yet obscured by the streamlined efficiency of the globalization period that has come crashing to an end.

Start with the economic concept of price elasticity, which is a fancy way of saying how much less of something people buy when its price goes up. An elasticity of −1 would mean a 10% price rise cuts consumption by 10%, one for one. Short-run price elasticity of oil demand averages −0.07 at the crude level, and China’s is −0.01 (EIA working paper on international demand elasticities, 2017, p. 13). At −0.07, crude goes up 10% and the world burns seven-tenths of one percent less of it. At −0.01, China burns one-tenth of one percent less. Read plainly: an importing country cannot dodge the first year of a price shock by using less. It has no mechanism to use less. It simply pays. The shock arrives as a current-account hit, a currency hit, an inflation impulse and — where fuel is subsidized — a budget hit, long before it arrives as any change in physical activity.

That second figure will look wrong to anyone following the trade press, where China’s crude imports have been falling and the obvious inference is that China is using less. Two different things are getting mixed together there. Elasticity measures consumption against price; imports measure purchases. A country sitting on 1.2 billion barrels of onshore storage can hold those two apart for a long stretch — buying extra when crude is cheap, buying less when it is dear, and burning roughly the same amount throughout. The stockpile is what turns a price shock into an inventory decision instead of a consumption decision. July made the point in public: with prices off their highs, Chinese crude imports jumped 22% in a single month to 8.45 million barrels a day, rebounding from a near-decade low without anything happening to the number of trucks on the road.

And where Chinese oil demand genuinely is falling, price is not the thing doing it. The electric fleet displacing gasoline and diesel was built for industrial-policy reasons and goes on displacing at $60 crude or at $110. That is substitution rather than elasticity, it runs on its own clock regardless of what the barrel costs, and it gets its own treatment at the end of this piece.

Diesel. The largest single cut, and the one that prices production rather than consumption. Freight, farm, mine, construction site, generator. Fuel was about 21% of the marginal cost of moving a truckload in the United States in 2024, at soft diesel prices. That cost is embedded in every physical object that moves: diesel on the farm, diesel to the silo, diesel to the shelf. And freight demand does not flex. The harvest — the few-week window when a crop has to come off the field or rot in it — cannot be asked to burn less. Governments facing a fuel crisis reach for work-from-home orders, speed limits and odd-even plate rationing — all of which target discretionary driving, because discretionary driving is the only demand a state can actually suppress.

Petrochemical feedstock. Two words are worth pinning down here, because they carry a lot of what follows. Naphtha is a liquid distilled out of a barrel of crude oil. Ethane is a gas — natural gas, the kind piped into a furnace, not the gasoline that goes in a car — and most of the ethane the chemical industry actually burns through is stripped out of natural gas at a processing plant rather than pulled off a barrel. Both end up in the same plants making the same things. They arrive priced off two different commodities, and that gap does real work later in this piece.

Together they run to roughly 12% of the barrel — naphtha, plus the share of LPG and ethane that goes into chemicals rather than onto a cooking ring — and that slice is the growth engine of oil demand while gasoline peaks and declines. That is not a forecast. Between 2019 and 2024, oil-based feedstock consumption rose about 2.3 million barrels a day, which is more than 95% of the entire net increase in world oil demand over those five years. In 2024 alone it was about three quarters of the growth. It lands on chemical-industry margins and on the price of everything made from plastic, fibre or fertilizer. Crackers can switch between feedstocks in weeks without touching their equipment, which lets them dodge which cut they buy and never the crude bill itself. This is also the one oil channel China cannot insulate, and I will come back to it.

LPG for cooking. About 10% of global oil demand, and the primary cooking fuel for 2.3 billion people in Asia. A crude shock here does not show up as a GDP decimal. It shows up as a subsidy line or a street protest.

Pay first, inflate second, slow third

That sequencing is the whole argument about timing, and it is what keeps this claim from overreaching.

Near-zero short-run elasticity cuts both ways. It maximizes the wealth transfer — the importer pays nearly the full increase, because it has no way not to. It also minimizes the immediate hit to activity, for exactly the same reason. Nothing switches off. An economy at $100 crude does not stall in the first quarter. It pays, its current account absorbs, its currency takes the strain, its inflation ticks, its subsidy bill grows, and the drag on real activity compounds over roughly four to eight quarters.

So the honest phrasing is margin compression and fiscal crowd-out inside the window, with the activity drag arriving on a lag. Expensive oil is friction the way a tax is friction. It does not shut anything down. It grinds margins, budgets and current accounts until something gives.

That also means that oil spikes do not do the work. Elasticity roughly doubles over the long run, from −0.07 to about −0.15, so the mechanism only reaches its second and third stages if the price stays up. 2022 was a spike — Brent was back near $80 by 2023 — and the machine never got past stage one.

Who pays, and why

The first piece argued that the damage runs through India, then Korea, then the European Union, then Japan, with China last, and that the sorting principle is legibility rather than alliance. Legibility is worth defining, since it does all the work here. It means how much of a country’s oil supply runs through channels Washington can see, price and interrupt — barrels crossing water the American Navy patrols, paid for in dollars that clear through American banks, insured in London, delivered on contracts that carry renewal dates. The question is never who signed which treaty. It is whether a switch exists and who can reach it.

Here are the magnitudes underneath that ranking. The arithmetic is mine, run on a sustained $30 increase against 2024–25 import volumes:

• India — 88.2% crude import dependence, a $137 billion import bill in FY25, and roughly $50–55 billion of added cost, about 1.4% of GDP. Each $10 on the barrel runs roughly half a percentage point onto consumer inflation and 30 to 40 basis points onto the current-account deficit.

• South Korea — around $30 billion, about 1.5% to 1.7% of GDP, the highest share in the group. Road transport is only about a third of Korean oil demand. The barrel goes to crackers and refineries, so a crude shock is a margin shock in the export-industrial complex.

• The European Union — roughly $100 billion, about 0.5% to 0.6% of GDP, landing through unusually diesel-heavy car and freight fleets, on top of a gas bill the bloc is already paying.

• Japan — roughly $35 billion, about 0.9% of GDP. Enormous dependence, but rich, holding 90-plus days of strategic reserve, and shrinking its barrel every year.

• China — roughly $120–125 billion, about 0.6% to 0.7% of GDP, against insulation that thickens every quarter: coal at 58% of power, electric vehicles past half of new sales, electric heavy trucks at 22% to 28% of new sales, 0.8 to 0.9 million barrels a day arriving overland by pipeline, and 1.206 billion barrels in onshore storage as of January 2026 — 104 days of net imports.

Half a percent to under 2% of GDP a year. Compounding tribute, and not a knockout. The narrow claim is the one that survives contact with anybody who checks: expensive crude is a recurring levy on the industrial margins and current accounts of America’s Asian and European client states, collected at precisely the moment those countries need capital for their own buildouts, while the United States — a net petroleum exporter of about 2.3 million barrels a day since 2020 — sits on the other side of the trade.

That last clause has a price attached to it. Brent is what a cargo costs on the water. West Texas Intermediate is what the same oil fetches in Oklahoma, and the two came apart precisely when the strait did. The gap between them averaged $11 a barrel through March 2026 and reached $25 on the last day of the month, the widest in more than five years. The Energy Information Administration’s own account of why the American benchmark lagged is domestic inventories and a scheduled release from the Strategic Petroleum Reserve. The war premium attached itself to the barrel that had to cross water. It did not attach to the barrel that stayed home.

The tier below

Everything above measures damage as a share of GDP. Go one tier down the income ladder and that metric stops working, because it is answering the wrong question.

For Pakistan, Egypt, Bangladesh, the Philippines, Turkey, Indonesia, Vietnam, Morocco and Thailand, the question is not what fraction of national output the oil bill consumes. It is whether the country can source the dollars to pay it. Sri Lanka’s reserves stood at $1.93 billion at the end of March 2022 against roughly $4 billion of debt falling due that year, and in April it suspended payment on its external debt rather than choose between bondholders and fuel cargoes. Pakistan restricted letters of credit through 2022 and 2023, and its goods imports fell 31% in one fiscal year — $80.13 billion down to $55.29 billion. Neither of those is a margin story. They are cannot-buy-it stories.

So the diagnostic down here is the oil bill measured against foreign-exchange reserves, against export earnings, and against whatever the state has promised to absorb at the pump.

Pakistan is the sharpest case. Its oil import bill for the fiscal year ending June 2026 came in at $16.86 billion. Reserves held by the State Bank of Pakistan stood at $17.04 billion on 31 July. The annual bill is roughly 99% of the central bank’s reserves. Import cover is 2.51 months and it moves by more than a billion dollars in a single debt-service week. The IMF program forbids the one domestic shock absorber available — relief on the petroleum levy was refused in April 2026 — so the pass-through lands on households and on the reserves line, with no fiscal cushion in between.

And the fragility axis turns out to be remittances. The three most exposed economies in this tier — Pakistan, Egypt, Bangladesh — share one structure: the external account balances only because money sent home by workers abroad rivals or exceeds total goods exports. Pakistan took in $41.6 billion of remittances in FY26 against $40.9 billion of goods and services exports. Egypt took $41.5 billion in 2025 against roughly $20 billion a year of non-oil merchandise exports. Bangladesh, $35.5 billion against $48 billion. Expensive oil in this tier is a claim on the diaspora’s paycheck. Which also means a Gulf labor-market shock and an oil shock are the same trade — and they are correlated.

The stress test already ran. This is what makes the section worth more than a forecast. Brent averaged around $103.70 in March 2026 and held near $100 into late July, before the August slide, so these countries were just tested at the levels the argument cares about, and their coping mechanisms are on the public record. West Texas Intermediate averaged about $91 across the same March. Pakistan and Thailand were not buying that barrel. Four types emerged:

• Pass it through, under IMF instruction. Pakistan and Bangladesh. The pump price is a tax machine or a monthly formula; the subsidy option is contractually unavailable.

• Freeze the price and eat the barrel. Egypt’s monthly petroleum bill went from $1.2 billion in January 2026 to $2.5 billion in March with pump prices held flat, and Cairo has committed to holding them flat through October. Indonesia had burned 203.7 trillion rupiah of a 315.7 trillion full-year subsidy-and-compensation budget by May, and 233 trillion by the end of June — roughly three quarters of a full year’s budget, gone in half a year.

• Bleed a stabilization fund. Thailand’s Oil Fuel Fund deficit reached 56.2 billion baht by early April, running roughly 1.5 billion baht a day on diesel compensation, with the fund applying for loans. It was still 58.4 billion baht underwater in mid-July, after crude had come off its highs. Vietnam’s fund went from about $213 million to about $7 million in a single quarter — exhausted — and the state budget now advances money into it. The shock became sovereign-adjacent debt.

• Suspend the market by decree. The Philippines, a deregulated fuel market since 1998, produced an energy emergency: mandated pump rollbacks carrying criminal penalties, the Senate preparing rationing, and crude supply partly running on an American sanctions waiver.

Nobody in this tier is Sri Lanka in 2022 today. Every one of them currently clears three months of import cover except Pakistan and Vietnam, and Vietnam’s oil bill is trivial against a $475 billion export machine. The fragility is real and it is conditional, and the condition is the thing to be precise about. A spike does not do this. One month of $120 crude on a headline does what 2022 did — the subsidy line stretches, the pain is real, and then the price comes back down and the machinery resets. What breaks a country in this tier is a floor that will not come down: quarters of it, long enough that the stabilization fund empties, the letters of credit tighten, and the reserves line stops recovering between cargoes. The level that breaks each one is set by its structure rather than its size.

Which is also where the strategy lives, and why I read the chokepoint pressure the way I do. A spike is an accident. A floor is a policy. My judgment on where each one goes underwater, measured in Brent and calibrated against what they actually did this spring: Pakistan and Egypt around $90 to $100 sustained; Turkey around $95 to $100, where the release valve is the lira rather than a default; Indonesia around $100 to $110, where the budget breaks before the balance of payments; Bangladesh around $110, where what breaks is the pricing formula rather than the reserves; Vietnam and Morocco around $120, the latter on price only, since with zero refineries and about 20 days of product stocks its real exposure is a diesel supply squeeze at any crude price.

And here is the part that complicates the frame rather than confirming it. This tier is where the oil weapon stops being friction and starts being coercion. But the three genuinely fragile countries are not rivals. Two of them are clients whose stability Washington actively underwrites through the IMF. Expensive oil down here does not starve a competitor. It raises the cost of holding the periphery together — the toll lands on the empire’s own maintenance bill.

That bill buys something, though, which is why the accounting does not come out simple. A client in fuel distress is a client that needs a program, and a program arrives with conditions attached. Pakistan is the worked example above: the levy relief it wanted in April 2026 was refused, and the refusal came from the lender rather than from any argument inside Pakistan.

Energy is a small lever to reach for when the ones already in hand are this large, though, and that is the thing to keep in view. Washington disciplines Europe and East Asia through dollar clearing, through defense procurement and basing, through commercial and technology linkages that touch every export sector those economies run. The oil bill is one more line in a ledger Seoul and Berlin were already reading. Nobody there needs to be taught what leverage looks like.

Which is why this tier is not really a story about who is a client and who is not. For most of the world economy the relationship with Washington is beside the point. The slowdown is the point. Sustained expensive crude will do three things at once, and all three run the same direction. It will put every competing industrial economy in a funk while the United States, a net exporter, has the room to run at its own reindustrialization. It will blow commodity prices out and then collapse them, and the solvent buyer holding the reserve currency gets to shop the wreckage for the metals its own buildout needs. And it will hand the energy layer itself — the half of it that moves by ship — to the country that already dominates that half. I have made this argument before. The oil transmission described in this piece is the machinery underneath it.

That is also the shape of the contest, and it is worth naming plainly. China is the electrostate — a term I did not coin and one that has been doing the rounds in energy writing for a couple of years. The United States is the molecule state. The usual counterpart offered is petrostate, and it does not fit: a petrostate lives on resource rents, and what Washington holds is not the rent. It is the shipping lane and the permission slip. And molecule is the right word rather than petro, for the reason this piece already spent a section on. What grows inside the barrel is the feedstock, and feedstock ends up in everything — plastic, fibre, fertilizer, solvent, pharmaceutical precursor. A country sitting astride those molecules is selling an input to industry itself rather than a tankful of energy, and that is a different kind of chokepoint.

One country holds the chokepoints on what moves by ship — crude, LNG, ethane, and the permission to buy any of them. The other holds them on what moves by wire and by battery — the cells, the panels, the processed minerals underneath both. Each is working to price the other’s dependence, and neither is anywhere near out from under the other’s grip. That contest is bigger than this piece and it will get one of its own.

The naphtha half

The gas piece took one half of the chemical seam. This is the other half, and it needs a short detour through what these plants actually are, because the entire argument sits inside them.

A steam cracker is the front door of the chemical industry. It takes a hydrocarbon — a molecule of hydrogen and carbon, which is what oil and natural gas both are — heats it past 800 degrees Celsius with steam, and breaks the big molecules into small ones. What comes out is mostly ethylene and propylene. Those two are the starting material for nearly everything synthetic: packaging, bottles, pipe, insulation, car interiors, textiles, adhesives, antifreeze, most of the fertilizer chain. If something in your house is plastic, it started in a cracker.

What you feed the cracker is the whole game, and there are two families of feed. Ethane is a gas, stripped out of natural gas at the processing plant, and its price tracks natural gas. Naphtha is a liquid, distilled out of a barrel of crude, and its price tracks crude. Same kind of plant, same products out the other end, different molecule in the front — and those two molecules are priced off two commodities that have come unstuck from each other.

American crackers run ethane. The United States has cheap domestic gas, and it is the only source of long-haul, ship-scale ethane on the water, so that advantage cannot be competed away by building a better plant somewhere else. It can only be rented. That was the gas piece’s argument.

Asian and European crackers run naphtha. When Brent goes up, their input cost goes up with it — and feedstock is roughly 60% of what it costs to run a cracker. The input is not a line in the budget. It is most of the budget.

Which leaves one question: can they charge more for what comes out? Only if somebody is short of ethylene and propylene, and nobody is.

Here is how the world got there. A world-scale cracker is a $5 to $10 billion decision with a construction schedule measured in years, so the plants coming online today were approved against demand forecasts written last decade. China approved a great many of them, for the reason it approves most things — to stop importing something it had been buying from abroad. The Gulf producers built their own, moving downstream to capture more value from a barrel than they get by selling the barrel. Both programs are landing at once, into a decade when plastic demand grows at its ordinary 2% to 3% a year.

That leaves capacity the market does not need, and a cracker handles that badly. It is a high-fixed-cost machine that wants to run flat out, and running it at three-quarters of capacity does not cost three-quarters as much. Any producer sitting on an idle unit will cut price to fill it, so the most desperate seller in the market sets the price everybody else gets. And nobody exits quickly, because a cracker is plumbed into the units downstream of it and is usually holding up a regional employment story besides. Overcapacity in this industry does not clear through a price signal. It clears through closures, one plant at a time, which is why the glut runs to about 2030.

So a plant facing a higher input cost and a buyer’s market for its output has nowhere to put the increase. It eats it. Expensive oil does not show up as expensive plastic in Seoul or Rotterdam. It shows up as a chemical business that stops making money, and then as a plant that closes.

Two caveats, because this claim is easy to over-drive and I would rather fence it myself.

Feedstock does not explain why European crackers specifically are in trouble. The decisive case is Grangemouth in Scotland — an 850,000-tonne-a-year ethane cracker, in Europe, running American feedstock, still losing money, and paying something like €100 million more for energy and up to €30 million more for carbon than an American equivalent in 2024. Operators name something else every time they are asked: energy prices at Teesside, lost integration at Antwerp, dead downstream customers at Terneuzen. The European fleet also averages about 45 years old, roughly twice the Middle East and three times China. European energy prices are doing the damage, and European carbon policy is second-order.

And the seam is compressing right now, not widening. Through late 2025 and into 2026, ethane crackers have faced rising feedstock costs while naphtha units benefited from cheaper crude. The structural tendency is for the gap to widen from both ends at once. That is a forecast, and the current print runs the other way.

What survives without qualification is the shape of it: gas-referenced molecules versus crude-referenced ones, and the advantage is transportable rather than territorial. Transportable is what turns it into an American chokepoint. An advantage built into your own territory — a better plant, a cheaper workforce, a shorter permitting queue — sits inside your borders, where no foreign government can touch it. An advantage that arrives by ship has to travel, and it travels through water the American Navy patrols, under insurance written in London, on paperwork Washington can refuse. Washington does not need to own the cracker at Antwerp. It only needs to sit astride the molecule going into it. Europe’s largest new cracker is being built at Antwerp specifically to run on imported American ethane. China took 47% of American ethane exports in 2024. Everyone else rents the American advantage; nobody replicates it. That is the tenancy structure from the first piece, showing up in a commodity almost nobody has heard of.

The permission channel

Everything above runs through price, and price is slow — four to eight quarters to the drag. There is a second channel that runs in weeks, and it is the one worth watching.

Washington does not control, nor does it need to control, all of the barrels of oil in the world. This is because it controls whether a barrel can be sold openly at full price to a named buyer. The chokepoint here is not a strait. It sits on the paperwork. Russian crude never left the market. What moves is the discount, and it moves with enforcement rather than with geology — $10 a barrel under Brent into Indian ports in January, a $7 to $8 premium at the height of the spring war, $2 to $3 under by August. Iranian crude clears through Malaysian paperwork, almost all of it bound for China — the tankers switch off their transponders at Malaysia’s Eastern Outer Port Limits and move the cargo ship to ship, sometimes more than once, to blur where the oil came from. The barrel flows. The value is taxed at the border of legality, and the rate is set in Washington.

The nuance that has to be got right: the discount accrues to the buyer, not to the United States. Sanctions are a transfer from Moscow and Tehran to Beijing and Delhi, and Washington takes no cut. What Washington holds is not a rent. It is an option. China’s cheap-crude advantage exists because enforcement is calibrated to permit it — ship-to-ship transfers in water the American Navy patrols and declines to interdict, paperwork nobody examines, insurance fictions nobody tests. Non-enforcement is the instrument. Tighten it and Beijing’s cost basis reprices without a single ship changing course.

That has always been the argument’s weakest evidentiary point, because it describes a posture rather than a document. It now has a document.

The Philippines imports Russian crude under a formal American sanctions waiver with a stated window — an initial term of 17 April to 16 May 2026, with an extension sought, renewed month to month. A treaty ally, 97% to 98% dependent on imported crude and products, with one refinery left since Shell closed Batangas in 2020, buys its marginal barrel at Washington’s monthly discretion. Not a metaphor about leverage. A renewal date.

The inverse case makes the same point from the other side. Indonesia’s celebrated Russian supply deal arrived at the worst possible moment: Russia asked roughly $9.50 a barrel above Brent for ESPO grade on June and July deliveries. Those barrels solve availability, not price. Any arithmetic that books discounted sanctioned crude as a durable structural advantage is booking an enforcement posture as a market fact.

Two clocks, then. The price channel is slow and grinding and lands on current accounts. The permission channel is fast and lands on a cost basis in a fortnight. The first is a tax. The second is a switch, and the hand on it is not in Beijing, Delhi or Manila.

What this does not claim

It doesn't turn off the lights. I settled this at the top, and it is worth restating because the wrong version of this argument keeps coming back. Transportation is heavily affected. But cutting the fuel cuts does not shrink the barrel neatly, it starves everything else that comes off the same crude — and those things are less optional than they sound. Aluminum is smelted using giant carbon blocks that burn away as the current runs, and those blocks are baked from refinery coke. That is the can, the car body, and the high-voltage line running to the house. Sulfur is stripped out of crude at the refinery and leaves as sulfuric acid, which is what dissolves phosphate rock into fertilizer. That is the price of bread. Nobody has ever built a refinery in order to make either one. They are leftovers, and leftovers cannot be ordered in larger quantities by wanting them.

Not a knockout. Half a percent to under 2% of GDP a year is a serious compounding tax on an industrial economy, and it is nothing like an embargo. The argument is stronger for conceding it.

Not fast. Everything downstream of sustained elevated crude is a forecast: oil going higher for longer, and not by accident, but by a deliberate set of policy choices combined with “accidental chaos” and defeat as a public relations veneer. In other words, cover.

What has actually been executed is the sanctions architecture, the rotation of China’s export book toward the countries in the tier above, and the peak-repayment schedule on Chinese lending to those same countries. The friction itself is positioned rather than realized.

And the timing element here is essential. Like stranded assets — the well or the pipeline that stops being worth anything before its working life is over — the timeline isn’t infinite. This is perhaps a general truism of chokepoints. Once executed, the clock of circumnavigation starts ticking.

This is because the oil-as-weapon depreciates — but on a schedule that somebody else sets. Electric vehicles displace about 5.4 million barrels a day by 2030 on the IEA’s own numbers, and China is the country doing most of the displacing — electric past half of new car sales, electric heavy trucks at 22% to 28% of new sales, and gasoil demand that likely peaked in 2022 or 2023. The country this argument treats as the hardest target is retiring its own exposure faster than anyone else, using an industry it already dominates. The friction window and the artificial-intelligence race window happen to coincide, and both of them are closing. The 2030 version of this argument is weaker than the 2026 version.

The two green transitions are worth separating, because only one of them touches any of this. Solar and wind compete in the power layer, i.e. against coal and gas. The opening section established that the power layer is not where oil lives, so a solar build-out — however much it reorders the electricity map — does close to nothing on its own to the friction layer described in this piece. Electric vehicles compete in the transport layer, and the transport layer is about 45% of the barrel. That is the transition the American oil weapon is competing against.

It lands unevenly, though, and the unevenness carries the argument. Set out the IEA’s product-by-product projection from 2024 to 2030 and the barrel splits into two halves moving in opposite directions: gasoline down 1.3 million barrels a day, fuel oil down 0.5, diesel and gasoil down only 0.3 — against naphtha up 1.1 and LPG and ethane up 1.9. Demand for oil as a combustible fuel may peak as early as 2027. Demand for oil as a feedstock peaks in nobody’s published forecast this decade.

Read that against the four channels and it says something narrower than “the weapon expires.” The transition drains the cuts that carry crude into households — the pump price, the subsidy line, the street protest. It leaves the freight cut nearly untouched and the cut that carries crude into industry growing. What depreciates is the political half of this weapon. What survives is the half that lands on chemical margins in Korea, Japan and Germany, and on the balance of payments of countries that import finished plastic and fertilizer. One thing to keep straight, since this is easy to overstate: feedstock is not booming. It’s just growing. It grows at a steady 2.1% a year and has become the dominant source of oil demand growth only because everything else stalled. It wins by standing still.

That relief gets counted in barrels, though, and barrels are the wrong unit for the tier that breaks first. Go back to Pakistan, Egypt and Bangladesh. None of those three fails because of the volume of oil it burns. They fail on the size of the bill, denominated in dollars and set against reserves denominated in dollars — Pakistan’s bill already running at roughly 98% of everything its central bank holds. Electrification takes barrels off their demand. It does not decide the invoice, because the price of the barrels still being bought is set somewhere else entirely. A country burning 10% fewer barrels at a price 20% higher is paying 8% more than it was, and has been relieved of nothing at all.

Which is also where the statistics mislead. Pakistan has been electrifying at a pace that makes for good headlines, and it is easy to read that as a country walking away from oil. Most of that build competes with the grid, and the grid is not what the oil bill is buying. Solar does take a bite out of the diesel burned in backup generators, and that bite is real. It does nothing about the trucks, the tractors and the cooking cylinders, which is the same list from the top of this piece. A country can rebuild its power sector completely and still owe nearly every dollar of that import bill.

The supply side runs the other way too. Producing fields decline at about 5.6% a year even with money still going into them, which is why close to 90% of global upstream investment since 2019 has bought standing still rather than growth. Set that against the demand forecasts, and the range is the useful part: from OPEC’s 122.9 million barrels a day in 2050 down to the oil majors’ own transition scenarios in the low eighties, every published projection has consumption moving less than a percent a year. The demand curve moves more slowly than the fields underneath it decline. That gap is filled by drilling, and drilling is a decision somebody makes annually.

And drilling is only one of the levers still in the drawer. Everything established above stays where it is: the permission layer deciding which barrels can be sold openly and to whom, the dollar the invoice is written in, the insurance, the lanes the cargo has to cross. None of those instruments is denominated in barrels, so none of them shrinks when the barrel count does. A smaller oil market still clears in dollars, still crosses the same water, and still runs on documents that can be withheld. The barrel count is what depreciates. The machinery built around the barrel does not.

The decline is uneven, which matters more than the average does. Non-OPEC conventional fields fall at 7.6% a year; Middle East supergiants at 1.8%. Whatever the barrel count looks like in 2035, a larger share of it sits in the Gulf. I have argued that the Middle East is already optional, structurally priced out by the shale revolution, and that holds for the United States as a supplier. It holds far less well for everyone still buying. The transition concentrates the price-setting function in the bloc with the slowest decline rates — the same bloc this argument has been treating as demoted.

Which leads to a cautionary tale, and it belongs at the front of the next argument rather than the end of this one.

A depreciating weapon argues for using it early, and using it early is exactly the reasoning that made 1914 and 1941 look like sound planning to the people who chose them.

Both of those are worth spelling out, because they are the whole warning. In 1914 the German general staff looked at the Russian rearmament program that French money was paying for, concluded the balance would be beyond recovery within a few years, and argued that a war fought now beat a war fought later on worse terms. In 1941 Japan sat under an American oil embargo with a finite reserve in its tanks, watched the fuel that made its navy mobile drain month by month, and reasoned its way to the same place. Strike while striking is still possible.

Both were internally perceived to be preventive wars, and both were argued by serious people whose arithmetic was broadly right. The windows really were closing. That is what makes the pattern dangerous rather than merely stupid — a closing window makes the reckless choice look like the responsible one.

The 1941 half is the sharper warning here, because the window closing on Tokyo was an oil embargo. The last time a great power ran an energy chokepoint against a major industrial rival, the rival did not fold. It went for the throat.

Editor’s note: I use Claude Code for data collection, analysis, mathematical modeling, drafting, and formatting (in varying degrees and times). High-level architecture is my own.

Stating this before the evidence so it can be held against me later. Oil stays higher for longer. American natural gas stays cheap. In geopolitical terms, they are two different machines, and only one of them is a weapon.

Why expensive oil is friction. Oil barely touches anyone’s electricity — it generates about 2.5% of world power, and the share has been falling for decades. Grids stopped burning oil after the 1970s shocks. That substitution already happened, which is why 2026 is not 1973 for the lights.

What crude still prices is diesel, which moves every physical object in an economy: the truck, the tractor, the mining shovel, the container ship. It prices the feedstock that chemical plants turn into plastics, packaging, synthetic fibre and fertilizer. And because no country can meaningfully cut its oil use inside a year, an importer facing a higher price simply pays it.

The scale is easy to underrate, because it arrives as a bill rather than as an event. If you run the arithmetic on a sustained $30 increase in the price of crude, against world consumption of about 103 million barrels a day, it moves roughly $1.1 trillion a year from the countries that buy oil to the countries that sell it. That is the transfer before anything in the global economy slows down.

How it plays out on the ground: Fuel was about 21% of the cost of moving a truckload in the United States in 2024, and that was at soft diesel prices. In India, every $10 on the barrel adds roughly half a percentage point to consumer inflation and 30 to 40 basis points to the current-account deficit. And when it gets bad enough the state eats it rather than the household: governments worldwide spent something like $900 billion shielding consumers from energy prices between early 2022 and April 2023. That is money not spent on anything else, during the exact years when every one of those countries needed capital for its own industrial buildout.

Expensive crude is a tax on making things and moving them, collected from everyone who has to buy the barrel — allies, non-aligned states and rivals alike, and it falls heaviest on those inside the tent.

And that is an inversion — or mean reversion, if you will — not a condition. Cheap oil is what made globalization physically possible in the first place. A supply chain that crosses three oceans only works when moving things is cheap enough that distance stops being a variable in the decision — the container ship, the air freight, the just-in-time inventory that assumed a truck was always minutes away. None of that was a law of nature. It was a price-setting regime, and it held for the 40 years of neoliberal globalization.

Turning that price back up runs the machine in reverse. And as I have been hammering on this Substack — and in lectures at Mines — I think it is being turned up deliberately.

There is a more comfortable story available, and it is everywhere right now. In that version Iran is a blunder: an administration that went in to denuclearize, got out over its skis, and is now improvising through a mess of its own making. Trumpian chaos. A historical mistake unfolding in real time.

That story requires nothing to be true except incompetence, and incompetence is always in stock. Mine requires something harder — that a policy consensus which survives changes of administration is executing a structural move off the globalization regime, and that the noise sitting on top of it is not the thing itself. I have named the specific institutions running it elsewhere, so this is not a claim about shadows.

Inference rather than a document. But the alternative is that every actor holding a lever happens to be pulling the same direction by accident.

Why American gas stays cheap, and which reason is doing the work. Two mechanisms.

Some American gas arrives as a byproduct of drilling for oil. You cannot pump crude without gas coming up alongside it, wanted or not. That remains an enormous share of total supply — but its growth is now halving, because the oil rigs that would produce it never came back.

At the extreme, the byproduct is worth less than nothing. In West Texas this spring the Waha hub averaged negative $3.81 per MMBtu and touched negative $9.60 — producers paying pipelines to haul the gas away. Nobody does that out of stupidity. The oil is the valuable product, flaring it off is restricted, and shutting the well to avoid the gas would mean giving up the barrel. So you pay someone to take it.

This condition will not last long, in my view. The AI buildout will largely run on natty gas, so expect more energy deals linking the data center buildout directly to the natural gas industrial complex. It is already happening at the wellhead: Chevron and Microsoft are advancing a 2.67 GW power project in West Texas under a 20-year agreement, and a campus near Odessa is being built to run behind the meter on Permian gas, with first power targeted for December 2026.

Note what that fixes and what it does not. These are off-grid builds sited on top of the stranded molecules, so they solve a West Texas plumbing problem — gas with nowhere to go — without touching the national price. Two or three projects of that size absorb something like a Bcf/d in the Permian. Against 111 Bcf/d of national production that is a rounding error. The negative basis goes away. The national price does not notice.

The rest comes from companies drilling for gas deliberately, and that is the part now setting the price. Gas-directed rigs are up about 25% on the year against 5% for oil rigs, and the fastest-growing gas basin in the country has essentially no oil in it.

So deliberate drilling is the driver today, not the oil byproduct. And if that reversed — if crude sat at $120 long enough that the oil rigs genuinely returned — the byproduct would surge again and push gas prices down, not up. Both roads arrive at cheap gas.

Why the cheap gas cannot escape to where prices are high. Natural gas is hard to export. It cannot be loaded onto a ship as gas; it has to be chilled to roughly minus 260 degrees Fahrenheit until it turns liquid, in a plant that costs $8–20 billion and takes years to build. America has only so many of those plants, and together they can handle about 17% of what the country produces. The other 83% has nowhere to go, so it clears at whatever domestic buyers will pay.

That clearing price has a name. Henry Hub is a pipeline junction in Louisiana, and the price of gas delivered there is the American benchmark — the number domestic contracts are written against, the number the futures market quotes, and the number this entire piece turns on. Europe’s equivalent is TTF, in the Netherlands. Asia’s is JKM. When the argument here says the spread never closes, those are the two prices Henry Hub is being measured against.

And the cheapness is the point. The gas is not being held down in order to sell it abroad. It is being held down because the electricity it generates is the input to the American artificial-intelligence buildout — the data centers, and the power plants going up to feed them. Every declared objective in Washington’s program needs that number low, and nobody holding a lever has any interest in raising it.

What that builds is domestic infrastructure, and domestic infrastructure is what reshoring runs on. The oil shock still ahead does the other half of the work — the more expensive it becomes to make things anywhere else, the better the argument for making them here. And access to the American AI stack will increasingly require physical presence: foreign multinationals, most of them headquartered in vassal states, moving operations inside the American border rather than buying the service from outside it.

I have made that argument at length in The Price of AI-Stack Admission. This is the energy layer underneath it.

The forecast graveyard

Let’s start with the scoreboard.

In February the Energy Information Administration had Henry Hub averaging $4.31 for 2026. By July that forecast was $3.67, with 2027 walked down to $3.49. The federal energy statisticians spent five months cutting their own number by nearly a dollar. The sell side made the same call and got the same result — that is the research desks at the big investment banks, the people paid to publish price forecasts for their clients. Through the winter they were calling for $4 and better, a few above $5. The summer market sat in the twos.

Now put that next to what happened on the demand side over the same five months, because every bullish story on the list actually came true. Feedgas — the pipeline gas arriving at an export plant to be chilled and loaded — ran roughly 17.9 to 18.1 Bcf/d in early July against about 13 Bcf/d in 2024. Golden Pass, a new export terminal on the Texas Gulf Coast, took its first cargo on April 22. Plaquemines, Venture Global’s plant in southern Louisiana, kept ramping. Cheniere brought more liquefaction trains online at Corpus Christi, a train being one processing unit — a terminal is built out of several, added one at a time. Gross LNG exports are tracking toward 16.3 Bcf/d for 2026 against 14.9 last year and 11.9 the year before.

Then demand showed up and the price went the other way.

Those forecasts were not wrong about demand. They just misunderstood who sets the price.

To borrow some trading lingo: betting that American natural gas is about to rise is called the widowmaker — market slang, and the joke is in the construction. It kills husbands.

It earned the name honestly. Strictly the word points at one specific bet, the gap between the March and April contracts, where winter draw-down ends and refilling season begins and a shifted weather forecast can move the price violently. In 2006 a hedge fund called Amaranth Advisors held that bet at enormous size. The gap collapsed by roughly 75% over two months, the margin calls ran past $2 billion, and the fund lost more than $6 billion and shut — the largest hedge fund failure of its era. Traders now use the word loosely for any wager that American gas is finally going up.

The reason it keeps killing people is the reason this year illustrates so cleanly: the demand story is usually correct, and the price still refuses to follow it.

83% of it cannot leave

Here is the whole mechanism, and everything else in this piece is detail on top of it.

The United States produces about 111 Bcf/d of dry gas. The export terminals can chill and ship roughly 18 of it. Call it 17%. The other 83% has nowhere to go. It cannot reach a world price, so it clears against domestic demand no matter what a cargo fetches in Osaka.

That is not a market condition. It is a plumbing limit, and it does not resolve the way a supply-demand imbalance resolves.

Build every terminal now under construction and the share of American gas that can reach a world price by 2030 still tops out somewhere around 20–25%, against a domestic pool heading past 115 Bcf/d. The tank stays roughly four times the straw for the rest of this decade.

Which is the real answer to why Henry Hub pricing does not converge upward to the European and Asian benchmarks the way a single global price would. The straw is too narrow to drain the tank. The spread narrows, but it never closes.

Supply is winning, and it does not need the oil price

The second thing the bull case underrates is where the gas is coming from now.

Dry gas production hit 111.2 Bcf/d in June, up 3.4% year over year, on a track to 115.3 by 2027. Gas-directed rigs stood at 126 in mid-July against roughly 101 a year earlier — about 25% growth, while oil rigs managed 5%. The Haynesville sits at 55 rigs with about 2 Bcf/d of spare pipeline capacity, and EIA projects it overtaking the Permian as the top growth basin in 2027. A basin with no oil in it, growing fastest.

That matters because it severs the gas price from the crude price in a way most people writing about this have not caught up to. The old story ran that expensive oil pulls rigs into the Permian, those wells throw off gas as a byproduct, and the byproduct floods the market. Every step of that still happens. It is simply no longer what delivers the outcome, because the market is being oversupplied by operators drilling for gas on purpose.

And the crude leg is weaker than it looks anyway. Horizontal oil rigs fell from 534 to 462 between January and April while WTI traded above $90 on 43 of 51 sessions, then turned back up in mid-June at lower prices. Against the spot price that sequence is incoherent.

Against the forward curve it is obvious. In June the Dallas Fed asked its panel of 125 firms how much production growth they foresee in 2027 if WTI holds at $100, $125, or $150 a barrel. The most-selected answer was the same bucket at all three prices, and the distribution did shift, with larger gains indicated at the higher prices. Operators are not indifferent to price. What that same panel does not believe is that the price stays: they put WTI at $78 two years out against a spot price averaging $87.27 while they were answering, and the survey’s index for expected capital spending next year came in at zero.

A drilling program is underwritten against the deferred contracts, not sold at the front month. A spike reprices the barrels already coming out of the ground. It does not finance the wells that would be drilled to chase it.

The data centers are real and they are late

The other half of the bull case is artificial intelligence, and the demand is genuine. The problem is when it arrives.

GE Vernova’s gas-turbine backlog runs past 100 GW and is effectively sold out through 2029, with pricing up roughly 300% in three years. Data-center gas burn adds 1 to 2 Bcf/d by 2027 and 4 to 6 by 2030. Most of the burn that the long-gas thesis is built on sits on the far side of a machine-shop bottleneck, and no amount of capital shortens a turbine queue that is already spoken for.

The wellhead projects described earlier are the exception that proves the rule. They are small, off-grid, and sited to solve a local stranding problem — which is why they can arrive in 2026. The grid-scale generation that would actually move national demand is the part standing in line.

The data-center burn is a 2029 event being argued about in 2026 headlines. It arrives — three years after the race it was supposed to decide.

The floor rose. The ratio did not move.

A concession is owed here, because the cheapest American molecules came from a condition that is expiring.

Those negative Waha prices came from a takeaway bottleneck rather than from infinite gas, and the bottleneck is being engineered out on a construction schedule. Kinder Morgan’s Gulf Coast Express expansion began moving volumes on June 9 and Waha recovered toward the flat line; Blackcomb should hold it there. The Permian operators themselves expect the constraint fully resolved in 2027.

The byproduct stream is thinning behind it. EIA has Permian gas growth halving into 2027, because the oil rigs never arrived. The marginal American molecule has migrated from free Permian byproduct to purpose-drilled Haynesville gas with a breakeven near $3.50. The floor moved up roughly a dollar.

Anyone pricing a wellhead should care about that. It does nothing to the argument here, because the level was never the claim. In July 2026 Henry Hub sat near $3 while Europe paid something like five times that and Asia more. A floor a dollar higher, measured against a European benchmark five times higher, is the same ratio.

A five-to-one price spread does not care where the one is.

What the cheap gas molecule actually buys

The most concrete thing cheap American gas purchases is not electricity. It is feedstock, and this is where the domestic advantage stops being a utility bill and becomes a physical input.

A steam cracker is the front door of the materials economy. It takes a hydrocarbon, heats it violently with steam until the long molecules snap into short ones, and produces ethylene, propylene, butadiene and benzene — the starting material for essentially all plastics, synthetic fibre, resin, packaging and a large share of industrial chemicals.

What it eats determines what it pays. American crackers run largely on ethane, which comes out of the natural gas stream and is priced against gas. Asian and European crackers run largely on naphtha, one of the liquids a refinery separates out of a barrel of crude, and priced against crude accordingly. This is not an analogy — it is how the market quotes the molecules. RBN Energy publishes an Ethane-to-Gas Ratio, setting ethane at Mont Belvieu — the Texas hub where American ethane trades — against Henry Hub at equal energy content, and it keeps that metric separate from its propane-to-crude and natural-gas-liquids-to-crude ratios.

So the American gas advantage and the global oil price meet inside one industry, from opposite ends.

The part that makes it strategic rather than merely fortunate is what happens when someone tries to buy their way in. The United States is the only country that exports ethane at long-haul, ship scale — Norway ships small short-sea volumes to two European crackers, and those crackers import American cargoes themselves. Nobody competes this advantage away by building a better plant. They can only rent it, by buying American molecules and shipping them home. INEOS is building Europe’s largest cracker at Antwerp to run on imported American ethane. China took 47% of American ethane exports in 2024.

And the landlord has already demonstrated the lease. In May 2025 the Commerce Department required a licence to export ethane to China. The last cargo left the day the rule landed. For roughly six weeks ships loaded and sailed under letters permitting everything except discharge at the far end, and the requirement was lifted on July 2. Per the American government’s own account, two Chinese cracker projects were delayed and may take naphtha instead — falling back to the crude-linked feedstock they built the plants to escape.

(Honest limits, since this is new ground. Feedstock is not why European crackers are closing — a European cracker running American ethane is still losing money, because 45-year-old plants and European power prices are doing that damage. And for 2025–26 specifically the ethane-to-naphtha gap is compressing rather than widening, as delivered ethane costs rise and naphtha crackers benefit from cheaper crude.)

The direct bid

The export contracts are where the domestic price stops being a domestic matter.

The standard structure is 115% of Henry Hub for the feedgas plus a fixed liquefaction fee, on a 20-year take-or-pay. Cheniere wrote that template and everyone since has copied it. The buyer’s signature is what lets a bank finance an $8–20 billion terminal, which means allied buyers pre-fund American infrastructure before a molecule ships. Then they import Washington’s domestic price as their own cost of doing business, and owe the fee whether or not they lift the cargo.

Roughly 40% of the world’s LNG contracts expire this decade — 92 of 245 by one count. Every expiry is a re-benchmarking event, where a buyer who had been paying a slope to Brent decides what to price the replacement against.

One caution, because this is where the argument usually gets oversold: adoption of the American formula does not climb a straight line. It moves violently with the cycle, taking better than 95% of new global volumes in the first half of 2022 and about 20% in the first half of 2025. Anyone describing a steady global conversion to Henry Hub pricing is reading one tooth of a sawtooth.

What grows, unevenly, is the pole. As of 2026 that is a claim about terms of trade and tribute rather than monetary architecture — the dollar’s plumbing is not at stake in an LNG contract, and the buyer’s margin is. That distinction describes the present rather than setting a ceiling. Each cycle of expiries adds to it, and the fee is fixed and take-or-pay, so it keeps arriving straight through the price collapse that wipes out everyone trading the volatile part.

Why everyone else’s gas stays expensive

The other half of the spread has almost nothing to do with American supply.

Crude still sets part of the foreign price, and the honest version of that share is a trajectory rather than a number. Oil-indexed volumes were 70% of the world’s contracted LNG in 2016 and the IEA has them at 48% by 2030; oil-linked pricing fell below half of global trade in 2024. Asia runs hotter than the global figure. The decay concentrates in expiries rather than renegotiations, at roughly 1.5 to 2 points a year.

But indexation is the decaying channel, and it is not what made Asian gas expensive this year. The live channel is simpler: a cargo that crosses an ocean carries a war-risk premium, a scarcity premium and an import-dependence premium, and a pipeline molecule running out of Louisiana carries none of the three.

The proof is that the Strait of Hormuz now outranks the index. An oil-linked LNG contract charges a set percentage of the Brent crude price — customarily 10 to 16%, a figure the trade calls the slope — so when crude moves, the contract moves with it. This year, buying a cargo on the Asian spot market has cost more than those contracts charge: high teens per MMBtu against roughly $12 on the formula, which works out to an implied slope near 22%. Buyers holding oil-linked contracts are paying less than everyone else. Hormuz is setting the Asian price, not crude.

The week of July 10 is that argument compressed into five days. Attacks on an LNG carrier near the Strait of Hormuz and escalating military tensions lifted the Asian and European benchmarks together. In the same five days, Henry Hub fell to a six-week low on soft weather and Freeport maintenance.

One commodity, two prices, opposite directions, same week. A single global gas market cannot produce that. Two engines can, and did.

I have made the broader case in Friction Is the Weapon and The Inversion of Abundance.

Two clocks, and why nobody competes it away

The spread runs on two clocks, and they get conflated constantly. Keeping them apart is most of understanding this market.

The cyclical clock is Hormuz. The spread widens when the strait degrades and compresses on the nominal reopening. Fully reversible, and it can re-widen inside a week — the July 10 tape is that clock made visible.

The secular clock is the buildout. Gulf Coast Express, Blackcomb and Hugh Brinson through 2027, then the export stack toward 2030, when the United States is set to supply roughly a third of global LNG even as three-quarters of its own gas still cannot leave. That one compresses in a single direction no matter what Iran does. Henry Hub grinds up toward the international benchmarks from below, they grind down toward it from above, and the gap closes from both ends.

The reason the spread survives anyway is that liquefaction arrives in multi-year, multi-billion-dollar blocks. Signed 20-year contracts gate when a train starts and steel gates when it finishes, and the run from investment decision to first cargo has averaged something like 3.5 years with a long tail past 7. Nobody competes that away by noticing the spread is wide. You cannot decide to be in this business next quarter.

There is a floor under it too. When the spread drops below the variable cost of shipping, the cargoes simply get cancelled — somewhere around a $4 Asian print at today’s Henry Hub.

None of the objectives are secret

Everything above describes a machine. This is the part where it stops being a coincidence that the machine produces what Washington has said out loud that it wants.

Three declared objectives, each with a document behind it and none requiring anyone to infer a motive.

Energy dominance. The executive order establishing the National Energy Dominance Council in February 2025 put the thing in the Executive Office of the President, chaired by Interior with Chris Wright’s Energy Department as vice-chair. Read the membership clause and the framing stops being rhetorical: the Council’s chair sits as a standing member of the National Security Council. Energy policy is formally wired into the national-security apparatus, by design and on paper.

Winning the AI race. The administration’s America’s AI Action Plan, released July 2025 under the banner of winning the race, runs to more than 90 federal policy actions across three pillars. One of those pillars is building American AI infrastructure. Strip the language and that pillar is power generation.

Denying China the frontier of compute. The January 2026 action on advanced computing chips, justified in its own title as protecting American economic and national security.

Those three converge on a single physical requirement: enormous quantities of cheap electricity, inside the United States, during a window that closes.

Now look at what the mechanisms in this piece actually deliver. Gas that physically cannot leave, so it clears domestically at a domestic price. A drilling program that keeps producing it whatever crude does. An export structure that ships the American price abroad on a 20-year lock while leaving the cheapness at home. And a political economy in which the vice-chair of the energy-dominance council runs the Energy Department, Scott Bessent’s Treasury wants the same number low, and every hyperscaler capital plan in the country is built on it staying there — with no organized interest anywhere pushing the other way.

That is not a conspiracy and it does not need to be one. It is a set of interests that coincide, sitting on top of a plumbing constraint that would produce the outcome regardless. The declared policy and the physical structure are pointed at the same place.

And the chip controls are the other half of it. Denying a rival the compute while feeding your own buildout the cheapest electrons on earth is one program with two instruments, not two programs that happen to rhyme. The export restriction handles the silicon. The straw handles the power.

(Tense marker: the objectives are documented and the mechanisms are executed. The claim that they are being run as one program is inference, and it is mine.)

What the contract knows

One last piece of evidence, and it is a document rather than an argument.

The American liquefaction business is not built to profit from the spread. It is built to be indifferent to it. The customer pays the formula plus a fixed fee under a 20-year take-or-pay, and the terminal is paid on volume whether or not the cargo is in the money. The delta between that formula price and the prevailing world price — call it $8–13/MMBtu at recent levels, $30–40 million on a single cargo — is captured by somebody else. It goes to the offtaker holding destination-flexible volumes, increasingly the large portfolio traders, whose share of contracted supply is climbing toward half by 2030.

A firm that expects a stable market signs for the market price. A firm that expects an unstable one signs for a fee and lets a counterparty absorb the variance. The contracts written in the middle of the last decade chose the fee.

Think of it as the juice — the cut a sportsbook builds into both sides of a bet so that it never has to care who wins. A standard point spread is priced at -110 either way: you lay $110 to win $100, whichever side you take, and the book keeps about $10 of every $220 wagered. Call it 4.8%. That is the entire business, and it is why the house does not need a view on the game.

The liquefaction fee is the same instrument and a much larger cut. Two to three dollars per MMBtu, fixed, on a cargo that lands in Asia somewhere around nine or ten, plus a 15% markup on the feedgas before the fee is even counted. The sportsbook keeps roughly a twentieth of the action. The terminal keeps something closer to a third. So in this case, the American house always wins, but it wins a lot more than Vegas.

The gap is the product, and holding the domestic floor down is how it gets manufactured: through export capacity too narrow to drain the domestic pool, through a drilling program that answers to its owners rather than to the price, and through a policy consensus that has quietly settled on the domestic number staying low. Energy runs 7–15% of a hyperscale data center’s total cost of ownership; servers are roughly 60%. Henry Hub would have to sustain something like $6–7 before fuel cost meaningfully bit the buildout, and nothing on the current strip goes there outside single winter months.

A $6 Henry Hub has no constituency. And the contracts were written to be indifferent to the spread, which tells you somebody expected a spread.

If you take nothing else from this, take the warning that comes with the name: beware the widowmaker.

Editor’s note: I use Claude Code for data collection, analysis, mathematical modeling, drafting, and formatting (in varying degrees and times). High-level architecture is my own.

The explanation on offer is demand, and it comes from people who know the market — USDA’s dairy analysts and the trade press have credited GLP-1 uptake and the protein boom for a year now. They’re right. Roughly 12% of Americans are on a GLP-1 and 70% say they’re trying to eat more protein. That is a real demand shock and it did move the price.

It’s also incomplete, and what’s missing decides who pays.

American cows are producing more milk than they ever have. Output in June 2026 ran 2.4% above June 2025, with production rising in 22 of the 24 reporting states and the national herd expanding. Cheese plants are opening. The USDA’s own outlook has cheese prices falling about 10% in 2026, to around $1.60 a pound.

Raw material abundant. Main product getting cheaper. The residue at an all-time high. That combination is the tell.

1. Nobody decides to make whey

Whey is what’s left in the vat when milk splits during cheesemaking. The curds become cheddar; the liquid is the leftover. You cannot build a whey plant and get more whey — you build a cheese plant, and whey falls out at whatever ratio the chemistry dictates.

That leftover is about 94% water and 6% solids. Everything downstream is a filtration decision. Dry it as it comes and you get sweet whey powder at 11 to 13% protein, sold under a dollar a pound. Push it through membranes and you get whey protein concentrate, or WPC, running 34% to 80%. Push it further and you get whey protein isolate, WPI, at 90% and up. That is the powder in your tub.

Here is the number that carries the argument. Of the roughly 200 million tonnes of liquid whey the world’s cheesemakers generate annually, on the order of 1% ends up as the high-protein grades. Wisconsin’s dairy extension analysts put it plainly: every whey product sits one filtration step away from another. What separates a 64-cent commodity from a 14-dollar ingredient is a membrane.

And every membrane in the country is full.

2. The bottleneck has a physical address

Membrane filtration systems and spray-drying towers cost hundreds of millions of dollars and take years to commission. There is no surge capacity and no third shift to add, because the third shift is already running. Rachel McGinness of Ecolab, which sells optimization services into these plants, describes membrane lines running 20-plus hours a day, with systems originally specified for 60% utilization now pinned at full. The efficiency hunt has descended to wash cycles: shave 45 minutes off a clean-in-place and you get one more vat through. When an industry is chasing throughput in wash-cycle minutes, the slack is gone.

Prices behaved accordingly. USDA’s Dairy Market News had isolate in the upper $13s to $14 a pound in July 2026, WPC 80% at $12.50 to $13.00, and buyers described as willing to pay whatever a load costs. Commodity dry whey — the leftover of the leftover — sat at 64 cents in late May 2026 against 52 cents a year earlier, holding above 60 cents continuously since September 2025. Across two years, WPC is up about 108% and isolate about 139%.

It now reaches the farm gate. Dry whey is an input to the Class III milk price — what a farmer gets paid for milk destined for cheese. Dairy analyst Betty Berning notes that every nickel on dry whey adds roughly 30 cents per hundredweight to Class III. Cheese isn’t paying the American dairy farmer this year. The residue is.

3. So who’s getting rich? Not who you’d think

The natural assumption is that somebody engineered this, or is at minimum sitting on it. The receipts run the other way, and it’s worth walking through them, because the answer relocates the blame rather than dissolving it.

Start with the most obvious suspect, and put the receipts on the table rather than paraphrasing them.

Glanbia owns Optimum Nutrition, BSN and Isopure — the tubs on the shelf, including the one this piece opened with — and it also manufactures whey ingredients. Vertically integrated, sitting on both ends of the trade. If anyone in this business were positioned to farm a shortage, it’s them. So here is what their 2025 full-year results actually report:

• Group EBITDA margin: 12.8%, down from 14.5%.

• Performance Nutrition — the branded division — 13.0%, down from 16.9%. A 390 basis point decline.

• The stated cause, in the company’s own words: record whey input costs.

• Performance Nutrition revenue +4.5%, of which volume was +3.6% — leaving under a single point of pricing.

Read that last line twice. Their input cost roughly doubled over two years and they put about 1% on the price, absorbing the rest into their own margin. Whatever else is happening here, the company that makes your protein powder is not the one profiting from what it costs. A business gouging you does not hand back 390 basis points to do it.

Now the genuine windfall, which is real and sits one step upstream.

For most of the history of cheesemaking, whey was a disposal problem. Plants sprayed it on farmers’ fields, discharged it into waterways, fed it to pigs, or used it to de-ice roads — the Parmesan producers of Emilia-Romagna fed theirs to the hogs that became prosciutto. A cost center with a hauling bill. The plants that installed membranes and dryers turned that liability into their best-margin product, and the arithmetic on the flip is worth stating precisely, because it’s the whole mechanism:

Cheese, the thing they set out to make, is forecast around $1.60 a pound in 2026. Whey protein isolate, the thing left in the vat, is running $13 to $14. The leftover sells for roughly nine times the product.

And the spread inside the leftover is wider still. Commodity dry whey — the same liquid, dried without the membrane step — sits at 64 cents. Isolate is about twenty-two times that price. Identical raw material, one filtration decision apart. That multiple is the entire prize, and it explains every capital decision in this industry right now.

Which is exactly why the withholding theory collapses. The payback period on whey processing investment runs under two years. Capital that returns itself in under two years does not sit in a drawer while somebody schemes. If the binding constraint were willingness, the lines would already exist — a 22x spread would have every processor in North America pouring concrete. They are pouring concrete. The constraint is engineering firms, stainless steel, and commissioning calendars, which is why every announced expansion carries a 2027 date rather than tonnage anyone can buy today. That build-out is real and it’s the light at the end of this particular tunnel; more on it below. None of it delivers a pound into 2026.

Two more facts kill the cartel reading outright. Beijing closed the door on this trade: more than 30% of American dry whey, WPC, isolate and permeate exports go to China, roughly 150 million of the 400 million pounds of dry whey shipped annually, most of it into piglet feed. In 2025 it took retaliatory duties past 80%, and the precedent was ugly — when tariffs bit in 2019 at lower rates, exports to China fell 55%. Losing your largest foreign customer is supposed to bury a price. It didn’t. Whey went sideways while butter, cheese and nonfat dry milk fell out of bed, and processors quietly stopped exporting the good stuff — WPC 80% export volumes fell 21% year-to-date through May 2026.

Record milk, a closed export market, a squeezed brand owner, and two-year paybacks nobody is declining. The shortage is an accident.

4. The allocation is not

Which brings us to the part that is somebody’s decision.

The demand story has three layers, and only the first two get discussed. Protein went mainstream — that 70% figure was 59% in 2022, and 40% of households now buy powder or shakes. GLP-1 drugs poured accelerant on it: 23% of U.S. households contain a user, and Circana projects those households will account for 35% of all retail food and beverage units by 2030. The drugs suppress appetite while making protein preservation a medical necessity, which is about as clean a demand curve for concentrated protein as anyone could draw up.

The third layer is where the price actually gets set, and it belongs to the marketing department.

Whey used to be bought by supplement companies. It is now bought by Nestlé, Danone, PepsiCo and Coca-Cola, bolting protein claims onto yogurt, bars, cereal, bread, granola, coffee and soda. And they buy on negotiated forward contracts — high-protein whey developed as a relationship business, and producers have sold forward well into 2026. The spot price you see quoted is the thin residual after the contracted volume is spoken for.

So look at what the contracted volume actually buys.

Everything has protein in it now. Protein soda — BellRing is putting Premier Protein Soda on shelves at 15 grams a can, and a brand called SkyPop is selling root beer and strawberry cotton candy on the same premise. Protein ice cream is this summer’s trend, and Protein Pints has moved into frozen novelties — ice cream bars dipped in chocolate and studded with quinoa, because apparently the shell had to pull its weight too. Cereal, chips, pretzels, cookies and pastries are all carrying 15-gram callouts. The trade press has stopped calling it a differentiator; the phrasing now is that protein is simply expected, in everything, as a baseline.

Protein potato chips. Protein cookies. If you want to eat protein, eat protein. There is no reason to smuggle it into your ice cream.

By this time next year I half expect to be drinking protein beers at the bar, which should make hitting my protein targets pretty easy. Except the market got there a decade ago — Barbell Brew has been selling a 21.8-gram protein beer since 2016.

The joke has a price tag, and here it is. British nutritionists have started calling the practice protein washing, and the side-by-side comparisons are the indictment. Protein granola: 15.8g per 100g, against 12.7g for ordinary granola. Protein bread: 20.3g against 13.4g. Protein yogurt: 9.4g against 6.5g for plain Greek. Marginal, every one — a premium paid for roughly three grams.

Then there’s the one that should end the conversation. A branded protein porridge delivering 10.9g per serving, against 15g for the regular porridge sitting next to it on the shelf. The protein version has less protein in it.

That product is consuming a globally constrained ingredient in order to carry a claim the cheaper product beside it already earns without saying so. The whey went into the label.

And that is where the tub money went. Every gram spray-dried into a cookie so the box can say 15g is a gram that never reached somebody using it as food — a lifter, a post-surgical patient, an eighty-year-old losing muscle mass, a GLP-1 user under actual medical instruction to hit a protein number. None of those buyers hold a forward contract. Frito-Lay does.

5. Nobody in the chain is motivated to fix it

Walk it, position by position. The cheese plants and dryers are building as fast as the engineering firms allow and profit either way — they were paying to spray this stuff on a field ten years ago. The ingredient processors have their best pricing in the history of the product. The packaged-goods companies pass the cost straight through, and the claim sells the box regardless of whether the gram count means anything. Glanbia is the one player in the chain the shortage genuinely cost — 390 basis points of margin — and it is building, with the New Mexico isolate line due in 2027. That is the point. The most motivated party in the sequence, holding the capital and the incentive and both ends of the trade, still cannot beat a commissioning calendar. So until the steel lands, it raises prices.

That leaves the person buying the tub, who holds no contract, bids for the residual, and has no representative anywhere in the sequence. Front-of-pack claims are not policed on whether the protein does anything, and no regulator anywhere treats a filtration bottleneck as its jurisdiction.

Which means the allocation corrects when the steel lands in 2027, on its own schedule, and not one quarter sooner.

6. The bad news, which lands first

Commodity input costs take 12 to 18 months to reach a shelf. The ingredient spike ran through 2025 and accelerated in early 2026, so another round of retail increases is queued for the first half of 2027 whether or not the ingredient market has already topped. The tub gets worse before it gets better, and that is baked in regardless of what happens next.

Expect the substitution playbook in roughly this order. Reformulation first — less isolate, more concentrate, blends cut with milk protein, collagen or pea. Collagen is worth naming: an incomplete protein that does not do the job whey does, prices well, and reads fine on a label. Then package shrink, the 5-pound tub becoming 4.65 and then 4 at a held price point. Then, if history holds, amino spiking — padding nitrogen content with free glycine or taurine so the lab assay clears while the usable protein doesn’t. That showed up the last time whey costs spiked. Third-party certification is worth more this year than it was three years ago.

7. The light at the end of the tunnel

It is real, it is funded, and it has dates on it. That last part is what separates this from the usual promises.

The relief comes as steel, and steel gets commissioned on a calendar. Glanbia’s New Mexico joint venture with Southwest Cheese adds roughly 4,500 tonnes of isolate capacity expected in 2027. Tirlán’s €126 million line runs mid-2027. Idaho Milk Products has $200 million going in. Coca-Cola is spending $650 million on Fairlife. Glanbia told investors in Q1 2026 that new global whey supply has started coming on stream and should continue through the year, with the caveat that demand is currently absorbing it as fast as it arrives.

That caveat is the honest version of the timeline. Capacity landing in 2027 means the ingredient market loosens in 2027 and the shelf follows 12 to 18 months behind. Call it late 2027 into 2028 before a tub gets cheaper, and that assumes demand merely stops accelerating rather than actually falling.

There is one early flicker on the demand side. May 2026 isolate production fell while cold-storage inventory built month over month. Rising stocks against falling output is what the top of a squeeze looks like. One month is not a call, but it is the first month in two years that pointed the right direction, and some analysts already argue the protein premium doesn’t survive past 2027 once the capacity and the demand curve cross.

Two more things could accelerate it. A plateau in GLP-1 adoption. And substitution succeeding at scale — if the packaged-goods buyers move their claims onto casein, soy or pea, they take their forward contracts with them and the whey stream comes free.

One thing that would make it worse, and it looks like good news. A China tariff settlement pulls American whey back out of the country and tightens the domestic market. If Washington and Beijing shake hands, the tub gets more expensive. Anyone reading a trade thaw as relief on this particular product has it backwards.

8. The company it keeps

Protein is expensive across the board, and the causes are unrelated. Beef and veal ran 11.8% above year-ago in June 2026 with USDA forecasting 10.7% for the year, on a liquidated cattle herd that has to get tighter before it loosens, because rebuilding means holding heifers back from slaughter. Ground beef hit a record $6.75 a pound in January. Eggs went the other way, forecast down 30.7% for 2026 as avian flu losses cleared.

Cattle is biology. Eggs are disease. Whey is capital equipment, and of the three it’s the only one where the pain is being distributed by contract rather than by nature.

I’ve spent time here on artificial scarcity — shortages designed into systems that have plenty — and on tenancy rather than ownership as the shape those arrangements take. This one didn’t need designing. The demand curve moved faster than a spray dryer can be poured, and everything after that followed from who had a signature on a forward contract and who didn’t.

The membranes are coming. They land in 2027, the price comes down, and none of that changes who got to decide where the scarce years went.

The scarcity was an accident. Spending it on granola was not.

Editor’s note: I use Claude Code for data collection, analysis, mathematical modeling, drafting, and formatting (in varying degrees and times). High-level architecture is my own.

Washington is trying to replace one economic order with another. The old one ran on the petrodollar and a financialized, high-margin, offshore-everything model. The intended replacement runs on physical capacity: energy, industrial plant, and artificial intelligence built on top of both.

Energy is the primitive in that plan. Everything else is a layer sitting on it.

The part worth arguing about is not whether this is happening. Capital is moving, contracts are being signed, and the policy consensus is visible in what nobody in Washington will touch. The part worth arguing about is the mechanism — what actually builds the American advantage, and who ends up carrying the cost of it. The answer to the second question is not mainly China.

One correction, and it is about size and scale

I have written about what I’ve dubbed ‘the Natty Greenback’ — American gas sold abroad at a formula tied to the American domestic price, paid in dollars, straight to American companies — as the thing that comes after the petrodollar. That framing invites a reading I want to shut down, because some readers took it and they were not being careless.

As a description of the tribute system, it holds up well so far since I dropped the Natty Greenback piece, and the rest of this piece defends it. Allies pre-fund American terminals, import Washington’s domestic price as their own cost, and owe the fee whether or not they take the cargo.

As a claim about the dollar, it is too small. Contracted American LNG fees run on the order of forty billion dollars a year. The reserve system they are supposed to be replacing is measured in trillions. Nothing at that scale substitutes for the petrodollar, and saying otherwise hands an easy kill to anyone who checks.

What it does instead is compound, and it is built better than the thing it is compared to. The fee is fixed and take-or-pay, so it keeps arriving straight through the price collapse that wipes out everyone trading the volatile part — roughly seventy-five billion a year by 2030 and ninety by 2040 on the current contract book.

The structural difference is the middleman. Under the petrodollar, the money went to Riyadh first, a rentier cut came off the top, and some fraction came back into Treasuries and New York banks. There was a foreign party in the middle getting rich on the transit. This arrangement has no such party. The buyer pays dollars to an American company, and that is the end of the sentence. Smaller than the petrodollar, and cleaner than it.

The dollar has never rested on one pole anyway. Capital-market depth, Treasury depth, the military, the courts, and now stablecoins all hold it up, and no single one of them is load-bearing on its own. This is one more, growing on locks that do not break.

Both things are true and should not be run together. A load-bearing pillar of the vassal system. A small, compounding, structurally superior pole of the dollar. The first is the argument this newsletter has been making. The second is a forecast about the 2040s.

One addendum before moving on, since it bears on the trade more than the argument. Look at where the drilling is actually happening: oil rigs have barely moved — up about 5% on the year, and they actually fell through three straight months of $90-plus crude — while gas-directed rigs are up about 25%. So even without oil prices rising and new oil wells throwing off natural gas as a byproduct, the gas is showing up anyway. Which means the so called ‘widowmaker’ trade (a reference to traders trying to speculate long on natty gas) looks to me to be still in effect. The floor under the price did move up about a dollar, since the marginal American molecule is now purpose-drilled Haynesville gas rather than free Permian byproduct — and the trade still doesn’t work. Consider yourself warned if you’re trying this lol.

Three machines, not one chain

Treating this as a single causal sequence is what makes it hard to follow. There are three separate mechanisms. They serve the same strategy and they run on different clocks.

Gas is the fuel. American gas is cheap because roughly eighty-three percent of it physically cannot leave the country — the export terminals can handle about seventeen percent, and the rest clears against domestic demand no matter what a cargo fetches abroad. Build every terminal now under construction and that share tops out near a fifth to a quarter by 2030. This is the input to the domestic buildout. It is fuel, not a weapon. (Executed and structural.)

Oil is the weapon, and it does not point only at rivals. Expensive crude does almost nothing to anyone’s electricity, because oil is about two and a half percent of world power generation and grids stopped burning it after the 1970s. What expensive oil taxes is the import bill, diesel, and petrochemical feedstock. And because nobody can use meaningfully less oil in the short run, the shock lands as a payment long before it lands as a slowdown: pay first, inflate second, slow third, with the real drag arriving four to eight quarters out.

The targets sort into three categories, and they should be kept distinct. China is the rival. Japan, South Korea and the European Union are vassals, in the specific sense that their security and their energy access both run through Washington. India is neither — an independent state that buys Russian crude, buys American arms, and takes instruction from nobody. The damage runs in almost the reverse of that order, which is the section below. (The channels are documented. The bite is forecast.)

Permission is the fast lever. Washington does not control the barrels. It controls whether a barrel can be sold openly, at full price, to a named buyer. Russian crude never left the market — it sells twenty to thirty percent under benchmark. Iranian crude clears through Malaysian paperwork. China’s cheap-crude advantage exists because enforcement is calibrated to permit it. Tighten enforcement and Beijing’s cost basis reprices without a single ship changing course. (Executed, repeatedly, and the fastest-acting of the three.)

That last one matters most for timing. The price channel takes years. The permission channel takes weeks.

The structure underneath all three

Nobody in this story is being denied energy. That is the thing that makes it hard to see.

What is happening is that access is being moved onto terms somebody else sets. The same shape appears in three different commodities:

In LNG, an ally signs a twenty-year take-or-pay contract at 115 percent of the American domestic price plus a fixed fee. Their signature is the collateral that finances the American terminal, and they import Washington’s domestic price as their own cost of doing business. The fee is owed whether or not they lift the cargo.

In petrochemical feedstock, American crackers run on ethane priced off domestic gas while Asian and European crackers run on naphtha priced off crude. The United States is the only source of long-haul, ship-scale ethane — so nobody competes that advantage away by building a better plant. They can only rent it, by buying American molecules. Europe’s largest new cracker is being built at Antwerp to run on them.

In crude, the discount a sanctioned barrel carries exists because enforcement permits it, and the rate is set in Washington. The discount accrues to the buyer rather than to the United States. What Washington holds is not a rent. It is an option.

Three commodities, one structure: tenancy rather than ownership. And every one of them moves by sea.

That is where the oldest layer of American power stops being background and becomes the enforcement mechanism. Alfred Thayer Mahan worked the argument out in 1890 and navies have been built on it ever since: sea power is not about winning battles at sea, it is about controlling the commercial routes and the chokepoints that commerce has no choice but to use. Britain ran that system for a century. The United States inherited it, and has spent eighty years underwriting the world’s shipping lanes.

A licensing requirement, a sanctions designation and a take-or-pay contract are all pieces of paper. What makes them enforceable is that the cargo has to cross water somebody else patrols.

The demonstration already happened, in a commodity almost nobody watches. In May 2025 the Commerce Department required a license to export ethane to China. The last cargo left the day the rule landed. For six weeks ships loaded and sailed under letters that permitted everything except discharge at the other end. The requirement was lifted on the second of July. Two Chinese cracker projects were delayed and, per the American government’s own account, may take naphtha instead — falling back to the crude-linked feedstock they had built the plants to escape.

Nobody was embargoed. A licensing desk decided, for six weeks, whether a molecule already on the water could be unloaded.

Who actually pays

This is the finding that reorders the argument, and it is the one I resisted longest.

Ranked by direct exposure to an oil shock, China is damaged least among the five major importers. India is first, at eighty-eight percent import dependence. Then South Korea, where the damage runs through petrochemical feedstock into industrial margins. Then Europe’s diesel-heavy fleets, then Japan. China lands at the bottom, and its insulation thickens every quarter — coal at fifty-eight percent of power, electric vehicles past half of new sales, heavy trucks electrifying, more than a hundred days of import cover.

That is not a failure of the thesis. It is the thesis. What makes an economy taxable is not hostility and not friendship — it is legibility. Open books, dollar invoicing, insurable cargoes, no shadow fleet, no tolerance for the paperwork fictions that move sanctioned barrels.

The vassals are legible by choice, because that is what the security arrangement costs. To be an enemy of America is dangerous; to be a vassal is fatal.

India is legible by necessity, which is a different thing and belongs in its own category. It is not anyone’s client. It buys Russian crude at a discount, buys American weapons, and takes instruction from neither capital. But it imports eighty-eight percent of its oil through open markets, and the discount that makes the bill survivable exists because Washington declines to enforce against it. That discount is not a right. It is a permission, and it can be withdrawn.

China is the one that has partially escaped legibility — through coal, through batteries, through pipelines, and through a shadow fleet the American government has chosen not to interdict.

Where the friction does reach China is indirect and slower: through its customers, whose purchasing power the toll erodes, and through petrochemical feedstock, the one segment of Chinese oil demand still growing and the one thing coal and batteries cannot insulate.

The honest magnitude across all of them is half a percent to under two percent of GDP a year. Compounding tribute, not a knockout. The argument is stronger for saying so.

Where this goes, and how confident I am

Three clocks are running, and they are not synchronized.

The burn-through. Scarcity cannot be manufactured on top of a glut, because a high price preserves the glut. The surplus has to be consumed first, which requires prices low enough to keep the world hooked on the barrel and to keep new supply from being drilled. So the sequence is long cheap, then short expensive. (Forecast. The tell is whether commercial inventories are still drawing, or the surplus is merely being repriced.)

The race window. Cheap domestic energy is an input-cost advantage during a specific window, and two things cap how much of it can be spent: gas turbines are effectively sold out through 2029, so most of the data-center burn lands after the race it was meant to decide; and fuel is only seven to fifteen percent of a data center’s total cost, so the price would have to roughly double before anyone in Washington cared. (Structural, and it cuts against the bullish version of the story as much as for it.)

The depreciation. The oil weapon is worth less every year. Electric vehicles and efficiency displace millions of barrels a day by 2030, and the Chinese petrochemical glut that suppresses margins worldwide has planned capacity collapsing after 2030. The friction window and the AI-race window happen to coincide, and both are closing. (This is the part that argues against waiting, and it is also the part that makes a preventive move dangerous — the same logic that made 1914 and 1941 look reasonable to the people who chose them.)

I am not going to pretend to a precision this system does not permit. Too many variables, too many discontinuities — a proxy strike, a suppressed damage report, and a policy reversal can all land inside a single week, and did. Things will zig or they will zag. The framework sits on top of any of the dated calls, and the calls get adjusted as things move.

What would falsify the framework

Worth stating plainly, because a thesis that cannot be falsified is not an argument.

If Henry Hub converges upward toward the international benchmarks while American export capacity stays a fifth of production, the plumbing claim is broken. If a second country begins exporting ship-scale ethane, the chokepoint claim narrows to a scale argument and eventually to nothing. If sustained elevated crude arrives and the deferred contracts follow it up without American drilling responding, the supply-discipline finding was a cyclical artifact. And if the friction shows up in Chinese activity data before it shows up in Indian, Korean and European margins, then the toll is not falling where I say it falls, and the legibility argument is wrong about its own target.

None of those has happened. Two of them are worth watching monthly.

Two pieces to follow. The first takes the gas side — why the American price stays low, what the export contracts actually do, and why the arbitrage is narrower and more durable than either the bulls or the bears have it. The second takes the oil side — how expensive crude physically reaches an importing economy, vassal or independent or rival, which is not the way most people assume, and what the permission channel can do that the price channel cannot.

Editor's note: I use Claude Code for data collection, analysis, mathematical modeling, drafting, and formatting (in varying degrees and times). High-level architecture is my own.

The Boulder Bulldog

The openly legal machinery that serves the few at the expense of the many — and the leverage it takes to change it.

By Evan Rowe

This is a spinoff of my foreign policy work. It’s semi related at times—but looks at the local level, and applies genesis architecture framing on the existing status quo in the American economy, particularly around rent seeking, effects on the broader majority (i.e. the bottom 80% of the class ladder), and other issues. Feel free to follow. The first major piece will be out this coming Monday, July 27th.

Read that way, the drone is Ukraine’s salvation — the mechanism by which a country running out of men substitutes silicon for flesh. It is the opposite. The same machine that kills at industrial scale is starving the one job it cannot automate: the man who stands on the ground and holds it.

Start with the physics, because the physics is not in dispute. A drone kills; it does not occupy. It can turn a supply road into a kill zone, but it cannot climb into a trench, clear it, and stay there through the night while the temperature drops and the next assault forms up. Holding terrain — the entire object of a ground war — still requires a human body in a fixed position, and the enemy still walks forward. For all the two million drones, Russian forces are larger than at any point since 2022 and continue to grind ahead. Mass production did not stop the line from moving. It was never built to.

Here is where the salvation story inverts. The drone did not lower the demand for infantry. It drained the supply of men willing to be infantry.

The International Institute for Strategic Studies lays the mechanism out flat: drone and one-way-attack teams “inflict most casualties and are perceived as safer assignments,” and so they “continue to recruit more personnel than infantry units. This translates to a small pool of infantry available for deployment.” A mobilized man, given a choice, wants the console, not the trench. He wants the unit that does the killing from two kilometers back — not the position where he becomes, in the phrase the soldiers themselves use, a bullet pillow. The technology is a magnet. It pulls every willing recruit toward the safe end of the lethality and leaves the dangerous end empty.

The receipts on that empty end are grim. Frontline brigades run at 50 to 60 percent of authorized strength, some as low as 30. Of a nominal force above a million, no more than 300,000 are engaged on the battlefront at all. Commanders describe sections of front five to ten kilometers wide held by as few as twelve men, average age forty-three. Ukraine mobilized roughly 200,000 soldiers in 2024 against a requirement near 300,000 — a target its own analysts called out of reach — while the pool of military-age men has fallen from 8.7 million before the invasion to about 5 million through death and emigration.

This is where the state’s own vocabulary gives the game away. The word now in official circulation is “busification” — Carnegie defines it as “the forcible detention of military-age men on the street and their transport by bus to enlistment centers.” Strip the coinage and it is a press gang: the eighteenth-century Royal Navy’s method for crewing ships no one would board on purpose. Complaints against the Territorial Recruitment Centers rose from 18 in 2022 to more than 6,000 in 2025. The Human Rights Ombudsman logged over 2,000 use-of-force complaints in the first half of 2025 alone. Zelensky has publicly ordered his officials to “deal with busification” — an odd instruction to issue about a program that officially does not exist. Defense Minister Fedorov concedes roughly two million draft dodgers and 200,000 absent without leave. Conscription now supplies about 70 percent of recruitment. The volunteer wave — disproportionately the nationalist believers who lined up in 2022 — ran out or was ground down in the early years. What fills the trench now is a coerced forty-three-year-old who was pulled off a sidewalk into a van—which has been met with a backlash that has largely been withheld from prominent media dissemination for years. For more in depth, and often ugly internal Ukraine power politics, I highly recommend Events In Ukraine, though to be clear, the higher echelons of the American foreign policy establishment are well aware of the press gang dilemma when it comes to manning the bullet pillow roles on the front lines, even if they sometimes, as in the case of Radio Free Europe/Radio Liberty, go to great propaganda lengths to try and frame the narrative as a clever Putin using propaganda to stir the unrest. I remain skeptical that your average 40 year old dishwasher in Ukraine requires a magic media bullet inserted in their head to prefer to not be kidnapped and sent to the front lines to die in conflict they have little stake in.

Put the two ends together and the paradox collapses into one mechanism. Every leap that makes the drone unit deadlier and safer makes the infantry role relatively worse — more lethal, lower status, harder to fill — so the modernization that reads as salvation cannibalizes the thing that actually holds the line. The machine war eats its own infantry.

There is a deeper reading, and it deserves stating precisely because it is contested. A Carnegie analysis co-authored by former defense minister Andriy Zagorodnyuk argues that Ukraine is not short of men in the first place — it “retains a substantial pool of military-age citizens,” enough to fill the ranks — and that the true failure is retention and command: converting bodies into trained, motivated, well-led soldiers and then keeping them. A fully equipped brigade rendered combat-ineffective by bad leadership; 200,000 men walking away. On that reading the manpower crisis is downstream of a leadership crisis, and the drones are a way of papering over both.

Both can be true at once. The trench is empty tonight for tactical reasons — the willing are flying drones and the unwilling are hiding — and empty this year for structural ones. Either way, the note every general staff from Washington to Beijing is quietly copying down is not that drones win wars. It is that drones change who has to die to hold the ground, and that a country can field the most advanced killing machine on the continent and still lose for the oldest reason there is: no one left who will stand in the hole.

Since January, Ukraine has been burning down Russia’s oil refineries. Long-range drones and cruise missiles hit Russian refining infrastructure at least 194 times in the first half of 2026 — about eleven times the rate of a year earlier, with a record month in May. Twenty-one of Russia’s thirty-eight large refineries have been struck since the start of 2025. At the peak last autumn, roughly a fifth of national refining capacity sat offline. The decisive front of this war is not in the Donbas. It is inside Russia’s energy system.

The weapons are Ukrainian. Fire Point, a startup that grew from twelve employees to about 3,700, builds the FP-5 Flamingo: more than 90 percent assembled in Ukraine, 3,000-kilometer range, a 1,150-kilogram warhead. The targeting is American. A missile with that reach still has to be told which refinery to fly into, and in October 2025 the Trump administration expanded intelligence-sharing specifically to help Kyiv hit “refineries, pipelines, power stations” — satellite imagery, monitoring data — and pushed NATO allies to feed the same stream.

Washington ran the proof itself. After the Oval Office clash in March 2025, it cut intelligence-sharing for about a month. Ukraine’s deep strikes went blind, and Russia retook eight settlements in Kursk within days. Feed off, the strikes miss. Feed on and widened, the refineries burn. Ukraine builds the spear; Washington decides where it lands. Europe writes the larger checks now — its cumulative aid passed America’s back in 2024 — but the layer that picks the targets is still American, and that is the layer that matters.

The target choice is the tell. Refineries make the diesel and gasoline Russia burns at home, so hitting them starves the Kremlin’s budget and lights fuel shortages at Russian pumps without pulling a barrel of export crude off the water. Taking Russian crude off the market would spike the global oil price and hand Moscow a windfall on every barrel it still ships. Striking the refineries does the opposite: Russia’s income falls while the world price holds. The budget shows it — oil-and-gas revenue down 34, 22, and 45 percent across recent readings, the federal deficit blown from a planned 1.2 trillion rubles to roughly 6 trillion, with the refinery strikes alone accounting for about a trillion of the hole. Gross export earnings, meanwhile, held up near €726mn a day in May, propped by a shadow fleet and a ruble that slid from 100 to 79. Moscow still earns hard currency abroad. It can no longer convert enough of it into the tax rubles that run the state.

That precision is where the Russia-Ukraine story joins a bigger one. Degrading a rival energy producer without moving the global crude price is exactly the calibration the United States is running across the whole board. Energy friction is the weapon, and firm-to-expensive oil is the goal, not the hazard: rivals get squeezed, importers slow down, and the United States — now a net energy exporter — collects on pressure it applies to everyone else. Russia is one target inside that design, and its punishment doubles as a market opening. Just like the Gulf energy system was once a vital center of American power—its energy assets are now energy competition—even if the sovereign wealth funds in the GCC are aligned.

The opening is gas, as in Natural Gas. The EU’s ban on Russian gas came into force on 18 March 2026, evicting Gazprom from the European market. Russian pipeline gas into the EU had already fallen from about 40 percent of supply in 2021 to 6 percent in 2025, and Europe is set to buy two-thirds of its LNG from the United States this year. A market Gazprom spent forty years building was pried open, and American export terminals were standing at the door. This is the Natty Greenback hatching into form — U.S. gas capturing the demand Russian gas used to hold, on long-term contracts that lock Europe onto American supply for the next decade. The war that degrades the competitor and reroutes the customer to the sponsor in one motion.

To be clear, none of this is “Ukraine winning the war” a refrain I’ve read and heard about from nepo baby bluecheck think tankers since the first 3 months of the war. (They’re about to turn this thing around guys!!) Russia keeps the manpower, the materiel, and the escalation dominance — the campaign that has left Kyiv on a few hours of power a day is the standing reminder that when Moscow decides to escalate, Ukraine is the one that absorbs it.

The proverbial Ukrainian “manpower shortage”, still ongoing, is no substitute for the much improved Ukranian drone force. As a Carnegie report — co-authored by former Defense Minister Zagorodnyuk — argues the deeper problem is conversion and retention, not demographic exhaustion:

Ukraine “retains a substantial pool of military-age citizens — enough to fill the armed forces’ ranks.” The failure is turning them into trained, motivated, well-led infantry.

So while the refinery strikes drain Russia’s war chest; they do not move the front. What they move is the energy map. A homemade Ukrainian weapon, aimed by American intelligence, is knocking down a major producer’s energy revenue and clearing Europe’s gas market for the United States — right as Washington leans into its own turn of the energy screw by leveraging Iranian retaliation in the strait—and I would forecast, eventually against GCC energy infrastructure, which will probably be the real upward spiral on energy costs—PRC buyer’s strike notwithstanding (I have no sense of what kind of depth the Chinese SPRs have… and frankly still more questions than answers there—but my base case remains that China reacted to the hormuz crisis as both a soft power and economic trading interest play for ASEAN states are and will be WAY more hurt by an energy shock than the PRC proper—but I remain open to future facts to change my assessment here).

Back to Ukraine—that is the shape of it in mid-2026. Ukraine burns the refineries. America aims the strikes and books the gas. Russia bleeds cash while it bombs the grid. The damage looks like Kyiv’s war; the beneficiary is the U.S. energy position.

The energy position is the setup, not the prize. The Russia dial is the setting anyone can see. The harder one is held in reserve, and it points at China — the rival Washington cannot bleed through oil the way it is bleeding Moscow, and the one racing the clock to get out of range. Sustained energy friction is built to do more than punish a producer. It is built to slow the world enough to put the whole commodity complex on sale.

Then the acquisition, where the shock loops back to the industrial rebuild at home. A deep enough contraction collapses not just oil but the whole commodity complex — copper, aluminum, the industrial metals a reindustrialization actually runs on — and the solvent buyer with the reserve currency gets to stock the buildout at fire-sale prices, sourced from wherever on the planet is forced to sell cheapest. The inputs for the North American rebuild trade on a global market; a global bust puts them on the clearance rack, and the buyer that stays liquid fills its warehouses and locks in long-horizon supply while everyone else is dumping inventory to raise cash.

The overall theme on here is, as it’s been for the past 6 months: The United States foreign policy apparatus is moving into a position of leveraging “energy dominance.” The shale revolution set the table; Venezuela, Iran, and Ukraine are the three acts. All three are presented as Trumpian mad king whimsical acts—but they all fall under a broader, well established, bipartisan set of foreign policy objectives, combined with a historical precedent: 1973 Yom Kippur War, The deliberate recession of the Volcker shock—as two cases in point. The internal public relations narrative is, and I predict, will become more net negative for the architects carrying the policy out, but will serve as a major power boost for core factional beneficiaries (i.e. the major players in the Genesis Architecture).

The energy scissors

The reflexive read on a Gulf oil shock is that Washington would move heaven and earth to prevent it, because a high energy price sounds like a tax on the American consumer and a headache for whoever’s in the White House. That read is a decade out of date, and it misses what an elevated energy price now does for U.S. strategy. It does three things at once, and they compound.

The first is energy dominance. The United States exports oil and gas rather than importing them, so a high price is revenue, not a wound. It widens the gap between expensive global crude and dirt-cheap domestic gas — Henry Hub under three dollars while Europe and Asia pay five times that — and every dollar of that gap shoves an industrial buyer in Rotterdam or Osaka toward American LNG. Expensive energy is a toll the U.S. now collects instead of pays.

And it isn’t only the gas side that wins. American crude producers do just fine with oil north of a hundred dollars, and this cycle they’re refusing to spoil it. Through the whole Hormuz-Iran scare the shale patch has stayed disciplined, holding output flat rather than drilling into the spike, because it got burned chasing the last boom into a glut and won’t do it again. The one force that used to cap every oil rally, American producers flooding the market the moment prices rose, has taken itself off the field.

And there’s a loop here that ties the oil price straight to the AI race, in Washington’s favor. A huge share of American gas isn’t drilled for on its own — it comes up as a byproduct of oil wells, especially in the Permian, where associated gas is now nearly half the basin’s output. Oil economics, not gas demand, decide whether those wells run. So a high oil price keeps the Permian pumping, and the gas it throws off floods a domestic market that can’t move it fast enough: Waha, the West Texas hub, has spent record stretches of 2026 at negative prices, producers paying to have the gas hauled away, and it kept crashing even as the Iran war spiked crude. That flood is what the AI buildout runs on. Data centers are electricity, that electricity is increasingly gas, and a high oil price quietly guarantees the gas stays cheap at the exact moment the buildout needs oceans of it. The same elevated crude that taxes every rival abroad subsidizes the American AI race at home.

The second is the global slowdown. High energy is friction, and friction slows every economy that has to buy its power abroad — China first, then Europe, then the Global South. The AI race is really an energy race wearing a lab coat: every model and data center is electricity converted into cognition at industrial scale, and the country running its buildout on cheap domestic gas sprints while the ones paying five-times-Henry-Hub for power stagger. You never have to touch a Chinese factory. You just raise the cost of the electricity that runs it, and the order books thin out on their own.

And China is uniquely exposed to exactly this pressure, because of how its model is built. Nobody beats China on price, but winning on price means operating on margins of three to five percent, and a country that thin has no cushion. Its national accounts don’t run on fat domestic profits; the surplus comes from export volume, from moving vast quantities of low-margin goods to the rest of the world. Slow that world down and you don’t dent a healthy margin, you erase a paper-thin one. The factories keep running. The orders stop coming. A demand contraction a high-margin economy would shrug off is close to existential for one whose entire buffer is the willingness of foreigners to keep buying.

The third is the one most people miss, and it’s the reason the first two aren’t just defensive. A slow global economy kills demand for the industrial commodities that build things — copper, AKA “Dr Copper”, the metal with a reputation for calling recessions before the data does, plus aluminum, steel, and the rest of the base materials — and their prices sag. Cheap inputs, though, only help a buyer with the means to buy: cheap domestic energy to run the smelters, a reserve currency that holds its value while others crumble, and a reshoring program hungry for exactly those metals. The United States has all three. Every rival that might also want the bargain metals is stuck in the mud — recession-bound, capital-starved, too underwater to reach for the goods floating past. So Washington can step in as a limited, selective buyer, feeding its reindustrialization at fire-sale prices, while the competition watches the sale it can’t attend. The friction that cheapened the metals is the same friction that keeps everyone else from using them.

Line those up and the shape is unmistakable. One lever — expensive energy — throws off export revenue, slows the rivals, and cheapens the industrial inputs that only the U.S. is positioned to absorb. Washington isn’t defending against an energy shock. It’s engineering a world of dear energy and cheap everything else, with itself as the one shopper left standing on the floor. It is engineering a recession it alone can afford to go bargain-hunting in.

Manufacturing Scarcity: Abundance is not acceptable. Artificial Scarcity is the order of the day.

As I discussed in The Inversion of Abundance, the missing geo-economics analyst gap in my view is the assumption material oil abundance will mean prices go ‘down and to the right’. But supply and demand in industry already showed a century ago, that when faced with competitive margin crushing competition, capital and the state leveraged artificial scarcity (and demand) techniques to prevent what the Chinese call ‘involution’, i.e. a competitive race to the bottom.

So… There’s a problem with the U.S. potentially wanting expensive energy, and it’s a big one. The world is drowning in oil. American shale is pumping around 13.5 million barrels a day, near record. Guyana came out of nowhere and is still ramping, Brazil’s pre-salt keeps climbing, Canada keeps grinding higher, and OPEC+ sits on millions of barrels a day of spare capacity it’s choosing not to pump. The government’s own forecasters expect the world to be oversupplied by more than two million barrels a day in 2026. Left to itself, a market this awash sags toward the cost of the marginal barrel, not away from it.

Which means the expensive half of the scissors doesn’t happen on its own. It has to be manufactured. The strategic question is not whether Washington wants energy dear — the three payoffs above answer that — but how you keep energy dear in the middle of a glut. That’s an engineering problem, and it has exactly two solutions. One is loud and bloody. One is quiet and negotiated. Both deliver the same macro.

Route A: make the scarcity real

Washington has treated Iran's ability to choke the Strait of Hormuz as its planning baseline for forty years — since Iran attacked Gulf shipping in the 1980s tanker war, through the Millennium Challenge war game of 2002, where a Red force modeled on Iran sank sixteen U.S. warships with swarming small boats in the opening days, to Caitlin Talmadge's 2008 assessment that mines and anti-ship missiles could shut the waterway for a month or more, to the Congressional Research Service spelling it out again in 2012. Anyone who has looked at the map has known the answer for a generation. The only variable was ever whether Tehran had a reason. So the idea that nobody anticipated this at the top is just beyond absurdity to me. I understand all the PR rationales that will come alongside this (Trump, Israel, et al), but i’m highly suspicious of those ideas.

So the U.S. can escalate Iran into hitting energy targets by continuing to ramp up pressure on Iran. Run the pressure on Tehran far enough and Iran reaches for the one target class it can always reach: Gulf energy infrastructure. Many analysts write this off because of the diplomacy between the GCC/Pakistan, and presumably China. But again, when the rubber meets the road, Iran can be left with few other options.

And spiking the price of gasoline to bloody an American president before the November midterms is an offensive play with a domestic-political payoff, and pipelines are soft. We have the rehearsal on tape. In September 2019, drones and cruise missiles hit Abqaiq and Khurais and knocked 5.7 million barrels a day — about half of Saudi output and 5% of the world's supply — offline in an afternoon

The thing to understand about Route A is why it works when a mere production cut wouldn’t. A voluntary cut is reversible; the market knows the barrels are still in the ground, so it prices a shrug. A bombing prices fear — the war premium, the money the market pays for the barrels it’s now terrified of losing next. Even Abqaiq only jumped Brent nearly 20% at the open before settling back and fading within weeks, because the Saudis reopened the taps fast. The version that sticks is the one that takes out what can’t be quickly rebuilt: the Hormuz escape valves. Saudi Arabia’s East-West pipeline, the Petroline, moves up to seven million barrels a day to the Red Sea, built in the 1980s to dodge exactly this kind of war; the UAE’s Habshan-to-Fujairah line lands crude past the strait. Those pipes are why a Hormuz scare fades. But the spare bypass capacity is thin, around 2.6 million barrels a day, and already run near flat-out. Put those lines under sustained fire and you don’t make fear — you make a stranded oil field, crude that physically cannot reach a tanker for as long as it takes to rebuild pipe under wartime conditions. That’s the one shortage an oversupplied world cannot backfill.

Route A has virtues, from Washington’s chair. The shock is real and durable. And it’s deniable: Iran wears the blame, the market reprices, and no one in Washington signs anything. Iran thinks it’s punishing the president; it’s delivering the friction the strategy wanted anyway — the instrument that mistakes itself for the agent.

It also has a defect. War is a blunt instrument, and blunt instruments overshoot. The same physical shortage that makes the shock stick can shove crude clean through the useful band into the uncontrollable one, and it flattens the infrastructure of the exact allies whose capital Washington is trying to recycle into its own industrial base. Whoever thinks they hold the dial is holding it in a burning room.

And there’s a subtler defect that changes the whole game. Route A only works while the Gulf stays a passive target, content to be the fuse in someone else’s device. But the GCC reads strategy as well as anyone. The moment Riyadh and Abu Dhabi grasp that their imperial patron is structurally content to watch their fields burn, the calculation flips. Washington’s short-term interest runs against them, whether by leaning on them to keep prices down for American drivers or by treating Abqaiq as an acceptable loss in a larger design, and a vassal that realizes it’s the mark stops waiting to be robbed. If a supply shock is coming either way, far better to author it than absorb it. That recognition is the door out of Route A, and it opens onto the second way to do this.

Everyone who walks through that door will need a story to tell their own public, and one is lying in plain sight. Global opinion of Israel has cratered; across three dozen countries a median of 67% now hold an unfavorable view, and the narrative that Israel dragged Washington into this war in the first place is already load-bearing across the region. So the cover writes itself: blame Israel for restarting it. Pin the war, the spike, the whole cascade on Israeli escalation, and every Arab government gets to move against the immediate interest of Washington’s closest ally while telling its own street it’s standing up to Israel rather than scheming with America. It’s a hypothetical, but a well-grounded one, because the narrative it rides was already built. And it has a second use: blaming Israel for lighting the fire keeps everyone’s eyes off the question of who actually directs whom. Everyone needs cover. Israel is the cover.

Route B: negotiate the scarcity

Here is the scenario worth mapping. Instead of waiting for Iran to blow a hole in the supply, the producers agree to hold the barrels back themselves — Saudi Arabia, the UAE, the rest of OPEC+, and, in the version that makes this interesting, Iran, all cutting in concert, with Washington’s quiet sign-off. A coordinated production strike dressed as a political rebellion against the U.S. and Israel. The GCC actors will have a few options, and given the information environment, propaganda seems to matter much less these days, as Trump himself has shown (people will probably dispute my sense of this). But they can simply say the U.S. has become unreliable, but the betting favorite for me is that Israel makes an ostensibly ‘rogue’ play, and the gulf states can finally say ‘no mas’ and offline goes the oil. It wouldn’t even be an OPEC+ story anymore because the UAE left already.

The bottom line: Oil strike without requiring Iran to blow up gulf infrastructure. When the U.S. energy friction play ends, they will be ready to be up and running—and regional diplomatic relations can continue on their impressive current path.

So here is the payoff: Because this is where the scenario earns its keep. Everyone at the table wins, each for a different reason:

The GCC wins by not getting bombed. Same elevated price, but with the plant standing and the revenue banked instead of craters at Abqaiq. Strictly better than Route A for Riyadh and Abu Dhabi.

Washington wins the entire scissors — energy dominance, global slowdown, cheap reshoring inputs — delivered bloodlessly and, crucially, on a dial. Where the bombing is a sledgehammer that spikes and overshoots and then fades, a coordinated cut is a thermostat: hold crude at a hundred-and-teens for as long as the strategy needs, adjust it quarter to quarter, and never flatten an ally’s export terminal to do it. For a sustained grind — which is what the macro actually wants, a long divergence rather than a one-day repricing — the dial beats the hammer.

Iran wins the sleeper prize, and it’s the biggest. It will probably bank victories already in hand, control over the strait, permanently removed sanctions, possibly some concessions in Bahrain, and continued improved economic ties in the region. In 5 years, none of the current energy dynamics will matter as it pertains to the Gulf, as any global negotiated settlement will still have an American footprint. Iran will probably not turn into a pre-1979 American client—but they very well may get out of the sanctions stranglehold they’ve been in since the revolution.

Why the negotiated version can actually hold

The obvious objection is the one that beat OPEC for a decade: cut in an oversupplied world and you just hand market share to whoever doesn’t cut, prices sag, and the cut leaks away. But the math has changed, and the supply side is the first reason. American shale, the swing producer that used to flood every rally, is holding output flat by choice this cycle, refusing to drill into the spike because it got burned chasing the last one into a glut. The escape hatch that used to defeat every OPEC cut is bolted shut from the inside, by the very producers who’d have to open it. A half-hearted trim still does nothing: OPEC+ took 2.2 million barrels a day off the table in late 2023 and crude barely flinched, but that was a cut with no discipline behind it and no slowdown ahead of it. The demand side is the second reason.

The strike doesn’t have to out-run supply forever, because it triggers the thing that shrinks the market underneath it. High energy prices push the importers toward recession, recession destroys demand, and global demand falls to meet the cut supply. The price holds without ever-deeper cuts because there are fewer buyers, not just fewer barrels. The friction becomes self-reinforcing: the cut causes the slowdown, and the slowdown validates the cut. That’s the same weaponized recession the whole strategy is trying to produce anyway — the cut supply and the destroyed demand are two ends of one operation. Weld them together and abundance stops being the escape hatch.

The cost you can’t negotiate away

Route B buys the GCC its infrastructure and buys Washington a controllable dial, but it charges a price the bombing doesn’t: deniability. When Iran craters a pipeline, Iran owns the shock and Washington keeps clean hands. When the producers hold barrels back by agreement, Washington is a co-author of expensive gasoline and the recession that follows, fingerprints on the thermostat. That’s survivable only if the strategic payoff outweighs the domestic-political hit — which, if the coalition running this treats congressional seats as cheaper than an industrial generation, it does. But it’s a real cost, and it sorts the two routes by timing. When deniability is precious — an election two months out — you let Iran do it kinetically and wear it. When deniability is affordable — after the midterms, in something like peace — you convene the quiet condominium and turn the dial.

And the linchpin is Iran’s sanctions relief: the single most politically explosive element in the American system, the one Tehran trusts least because it has watched relief get revoked before, and the one that strains hardest against Iran’s own grievance about suppressed exports the moment you ask it to cap output. The whole negotiated route lives or dies on whether that relief can be made to look permanent enough for Tehran to sell flat and cheap-ish rather than pump flat out.

The world it produces

Both roads arrive at the same country. Scarce energy, cheap everything else, and one economy — insulated by its own oil and gas, its own currency, and its own re-shored factories — walking the aisles of a fire sale it lit, while the rivals stand in the recession that lowered the prices, unable to reach the shelf.

The warhead and the handshake are the same weapon pointed at the same target. The only question Washington actually has to answer is how much cover it needs the day it pulls the trigger.

The United States is sitting on one of the largest energy gluts in its history. Natural gas at the American benchmark, Henry Hub, trades around three dollars per million BTUs; out in the Permian Basin of West Texas there’s so much of it, and so little pipeline to move it, that producers have spent stretches of 2026 paying to give it away at negative prices. Physically, the country is drowning in cheap energy.

Here’s the forecast. Almost none of that abundance is going to reach the rest of the world, because Washington is building the machinery to keep it from arriving.

The pieces already on the board are real. American gas, chilled to liquid and shipped abroad as LNG — liquefied natural gas — lands in Asia and Europe at roughly four to seven times the domestic price, and companies like Cheniere and Venture Global pocket the spread. A 2025 law, the GENIUS Act, routes the dollars those exports throw off back into U.S. government debt; Treasury Secretary Scott Bessent has said out loud that it will “lead to a surge in demand for US Treasuries.” Energy Secretary Chris Wright runs the export-licensing levers. The terminals keep getting bigger.

What those pieces are being built toward is the part I’m forecasting, because it isn’t finished and the people running it don’t narrate it. The bet is to hold the world’s energy expensive while America’s stays cheap, and to leverage that gap as friction — a deliberate drag on every economy that has to export to survive. Germany’s chemical industry, cut off from the cheap Russian gas it was built on, is already losing ground to American rivals. China imports little energy itself, but its customers do, and a Europe paying triple for power buys fewer Chinese goods. Keep global energy costs elevated and the export economies slow down. On the forecast, that slowdown is the whole point.

This would be a brand-new idea if it were new. It isn’t.

Industrial capitalism has a recurring problem: it gets too good at producing things. Productive capacity outruns what people can buy at prices that cover the producers’ debts, prices start to fall, and falling prices threaten everyone who borrowed against the old ones. The fixes have a hundred-year track record, and almost none of them involve making more for less.

The advertising industry was one fix. Edward Bernays, who built the field, wrote it down in 1928: a modern factory “potentially capable of supplying a whole continent with its particular product, cannot afford to wait until the public asks for its product.” The factory can’t wait, so the system manufactures the wanting. Demand engineering exists to soak up an oversupply that would otherwise crater the price.

Planned obsolescence was another. In 1932 a Manhattan real-estate man named Bernard London proposed that the government legally assign every product a death date and tax anyone who kept using something past it. London’s version was too blunt to pass, but the soft version became standard practice — Brooks Stevens defined it in 1954 as “instilling in the buyer the desire to own something a little newer, a little better, a little sooner than is necessary,” and General Motors built it into the annual model change. Shorten the life of the thing and you’ve manufactured a shortage of the current version.

Then there’s branding, which can make a common thing scarce by fiat. De Beers sat on a mountain of South African diamonds — a stone that is not actually rare — and spent forty years turning cheap carbon into the thing a groom was expected to spend two months’ salary on. Restrict the abundant good, sell the restriction, collect the margin.

The government, and the then American left New Deal Coalition got into the business openly in the 1930s, when the market’s own machinery wasn’t enough. The New Deal paid farmers to kill six million pigs and plow ten million acres of cotton under while breadlines ran around the block, and the National Recovery Administration wrote five hundred industrial codes fixing prices and capping output to kill what its own administrators called “destructive competition” — the kind that lets efficient producers cut prices until the weak ones die. For those following me closely see Crush margins or die trying and The British read from this script first on the classic deflation/inflation discussion from a historical policy standpoint.

And the move never stopped. AbbVie wrapped its best-selling drug in 311 patents to keep cheaper copies off the U.S. market for seven extra years. Apple takes a 30% cut of the app economy it controls. Hermès makes its best customers qualify to buy a handbag. Different decade, different industry, same maneuver: face an abundance that threatens the margin, restrict the supply or manufacture the demand, and keep the price from clearing.

The reason this looks novel is that a lot of people came up during the one stretch when it was hard to see. From roughly 1980 to 2020, the American experience was cheap imported goods and rising asset prices, and it looked like markets had finally been left alone. They hadn’t. The margin management had just moved offshore and into finance, to layers most people never watch. A whole generation mistook that exception for the natural state of things, what is coming next is a return to the old.

The energy play is the newest entry in a very old ledger. The commodity is bigger this time — gas and oil instead of cotton or carbon — and the target is the industrial base of every country that has to export to compete. But the logic is the one the cotton farmers knew in 1933. Abundance is the threat. Scarcity is the product. And the cheapest energy on earth does the most damage when almost no one else is allowed to have it. Oil and gas abundance analysts are overlooking the politics to view the economic reality: Tons of hydrocarbons=abundance. But abundance rests in a political history that American elites have long sought to solve in order to maintain the wealth hierarchy, and prevent ‘ruinous competition’ from eating into the margins of sectors controlled by certain segments of elite power. What I believe is coming next on the energy front will be similar—but initially established by the national security state in the U.S.

The U.S. built a ‘globalization’ regime 30 years ago, it set the rules and won, the benefits. But China entered INTO that system and beat Washington at its own game. Now it’s picking up its ball and going home.

Everyone calls it the AI race. Under ordinary circumstances, a race has exactly one rule — first across the line wins — and the better part of a trillion dollars Washington is pouring into data centers and power plants says somebody in charge believes that rule holds. The U.S. dumping a gazillion dollars into the AI Capex does not indicate anything resembling SLOW.

Yet, three days after Anthropic shipped the best model in the world, the Commerce Department sent a letter at 5:21 on a Friday and made the company switch it off. The country supposedly sprinting hardest reached over and pulled its own fastest runner off the track.

That’s a paradox only if it’s the race everyone thinks it is. It isn’t. But I don’t think Washington isn’t trying to win by finishing first — it’s trying to change the rules of the race.

Let’s do a little sports analogy: In 2004 the Detroit Pistons beat the Los Angeles Lakers for the championship, and on paper it shouldn’t have been close. That Lakers team had four future Hall of Famers — Shaq, Kobe, Karl Malone, Gary Payton — the most loaded roster money could assemble, the superteam, a bleeding edge of basketball talent. Detroit had no superstar. What it had was the best defense in the league and a refusal to play the game L.A. wanted to play. The Pistons dragged every possession into the half-court, made it ugly and physical and slow, and beat the superteam 4–1. They didn’t win by being more talented. Nobody was more talented than that Lakers roster. They won by changing what kind of game was being played until raw talent stopped being the thing that decided it. That’s the whole American strategy in one series, and it’s worth understanding why, because there are two kinds of games here. American policy planners don’t see the traditional first mover advantage as an advantage now. The entire play is dragging this contest into a slow grind.

Two kinds of games

Some contests are won on speed. Whoever moves fastest, copies fastest, scales fastest, and drives the price down fastest takes the field. Open information, low barriers, quick diffusion — the advantage flows to the quickest study and the cheapest producer.

Other contests are won on friction. Whoever controls the gates, the chokepoints, the scarce inputs everyone else needs — they win by deciding who gets through and on what terms. Speed doesn’t help you if the thing you need to go fast is held by someone who can simply say no.

China wins at speed. That is its entire structural advantage: manufacturing scale, deployment velocity, a state that floods a sector with cheap capital until margins hit zero and the foreign competition breaks, and a culture of copy-improve-undercut that turns any open technology into a commodity within a year. Give China an open field and a fast clock and it wins almost every time. That’s the model they built. Scale and speed. We watched it happen in solar, in steel, in batteries, in electric vehicles, and over the last eighteen months, in AI models.

The United States now views it’s leverage as one where friction is the tool. The old order of just in time production, and complex global supply chains is done. The U.S. advantages are all gatekeeping advantages: the deepest capital markets on earth, the dollar and the payment rails that clear through it, abundant domestic energy, the U.S> Navy sitting on every maritime chokepoint that matters, and a lock on the handful of irreplaceable nodes in the chip supply chain — the lithography machines only the Dutch (ASML) can build, the design tools, the leading-edge fabs. America doesn’t win by being fastest. It wins by owning the gates and charging admission, or denying it.

This is the lens for the entire strategy I keep writing about. Strip the Genesis Architecture down to one sentence and it’s this: convert speed games into friction games. Take every contest where China’s velocity wins and drag it onto terrain where America’s control of the gates wins instead. Energy, finance, chips, shipping — the whole program is a campaign to make the world’s contests turn on chokepoints rather than clock speed, because chokepoints are the thing the United States still owns outright. The Iran war is just a minor data point in all this, another instrument to slow the whole global operation down, but other examples are now abundantly clear.

Energy is the template

The clearest version of this move isn’t AI. It’s energy, and Washington has been running it the longest.

Cheap, abundant energy is a speed game. When power is cheap everywhere, the advantage runs to whoever builds and manufactures fastest — which is China. A world awash in cheap energy runs at China’s tempo. Expensive, gated energy is a friction game, and a friction game has a referee: whoever holds the cheapest supply and controls the flows everyone else depends on. That’s the United States — the swing supplier of liquefied natural gas, sitting on cheap domestic gas priced off its own Henry Hub benchmark while its exports help set the price the rest of the world pays. So when Doomberg repeats a mantra: "The long-term real price of all commodities is lower." it’s economically correct, but misses what I view the emerging geopolitical reality. In a sense, this is a truism of more than a century. Natural capitalist production modes lend to deflationary pricing via competition.

So under normal global trade conditions, then yes, the long-term real price of all commodities is lower. They’re right. Under normal competition, that’s exactly what happens — production gets more efficient, supply expands, and the real price of stuff grinds down, decade after decade. For most of the globalization era that was the whole story: commodities were going to get cheaper, forever, and the smart money just had to be patient.

But that’s economics talking, and economics no longer has the floor. The defining feature of the world now taking shape is artificial scarcity, imposed on purpose, by state power — sanctions, chokepoints, export bans, security guarantees handed out and withheld. Washington has discovered it can override the natural price of a thing simply by deciding who’s allowed to buy it. The deflation Doomberg correctly predicts is the deflation geopolitics now exists to prevent. The price wants to fall; the state won’t let it. Geopolitics trumps economics, and it has for longer than most people remember, the last 30 years will look more like an aberration, than the norm going forward.

Washington has a structural interest in a high-energy-cost world, as counterintuitive as that sounds coming from the country that runs on cheap gas (again—why I have argued from the early Hormuz closure: “If Iran didn’t close the strait of Hormuz, the U.S. Navy would be tasked to do it”.

The trick is that the two aren’t in tension: cheap at home, scarcity imposed abroad. Elevated global energy prices drain the surpluses of the manufacturing economies that have to import the stuff — China and Germany first in line — while the U.S. keeps its own input cheap and collects a toll on everyone else’s. The friction slows the whole world down. It just doesn’t slow it down evenly. It slows everyone else more, and it hands the United States the one prize that matters most in a friction game: pacing control. The power to set the tempo — to decide, by widening or pinching the export valve, how fast the rest of the world is allowed to run.

That is what “energy dominance” actually means. Not energy independence, not merely having a lot of the stuff. It means owning the dial. And once you’ve watched the move work in energy — slow the whole game, keep your own costs low, hold the tempo — you recognize it on sight when it shows up somewhere new. Like AI.

AI is a speed game by nature

Now look at artificial intelligence with that lens, and the problem is obvious.

AI, left to its own nature, is the purest speed game ever invented. The core product is a file of numbers that copies for free. A frontier capability, once it exists, diffuses outward in months — through open-weight releases, through distillation, through researchers who read the published paper and rebuild the result, through talent that walks from one lab to the next carrying the method in its head. There is no physical chokepoint on a model the way there’s a chokepoint on a barrel of oil or a leading-edge wafer. The thing wants to spread. Spreading is what it does.

Which means that AI, in its natural race state, is played on China’s home court. Fast, open, cheap to copy, brutal on margins. I spent the last two posts documenting exactly that: Chinese open models a few points off the frontier at a twentieth of the price, the edge diffusing to everyone by Christmas (or pick your next months away holiday by the time you read this), the American labs getting commoditized into utilities. That’s not an accident or a failure. That’s just what a speed game does, and China is better at speed games than anyone alive.

So if you’re Washington, and your structural strength is friction, and the most important technology of the century is natively a speed game your rival dominates — what do you do?

You change the format. You drag AI off speed-turf and onto friction-turf. You take a technology that wants to diffuse freely and you wrap it in gates: export licenses, chip controls, an imperial network attached to Pax Silica that continues to tether alliance states by bonding them to the U.S. compute engine, foreign-access bans, the power to pull a model off the market with a letter. You turn intelligence from something that spreads into something that’s permissioned. You make it behave like oil through a strait instead of code on a hard drive.

That is what the Fable pull was. Not a safety measure — the pretext was paper-thin, and Anthropic said so out loud. The legal machinery is worth naming precisely, because it tells you how Washington is now thinking. The order barred foreign nationals from the model, which invokes a doctrine called the “deemed export” — the rule that releasing controlled technology to a foreigner, even one standing on American soil, counts as exporting it to their home country. That doctrine exists for munitions and nuclear-weapons design. Applying it to a commercial chatbot is the tell. Washington just classified frontier intelligence as something closer to an armament than a software product — a thing whose diffusion is a national-security event, to be licensed and gated like a weapon. The friction regime didn’t grow around AI by accident. It’s being built, deliberately, lever by lever.

Why slowing down is the move

Here’s the part that makes this a strategic offensive maneuver rather than reactionary, and it’s the argument the last piece ended on.

If an AI bleeding edge doesn’t last — if whatever you build at the frontier diffuses to your rival in six months — then racing flat-out at the frontier is a sucker’s move. I’ve been describing this as the second mover advantage. You pay the full cost of pioneering, and China pays a twentieth of that to copy the result. Sprinting ahead in a speed game your opponent will catch up in two quarters isn’t winning. It’s running an unpaid research lab that the PRC can leverage into its own advantage.

Once you accept that the frontier leaks, the optimal play flips. You stop trying to be permanently ahead, because permanently ahead isn’t available. You start trying to control when the capability lands and who gets it. The winning move in a leaky game isn’t to run faster. It’s to install the gate and become the one who decides what passes through it. Stop being the leader. Become the toll collector.

This is why “slow down” is the wrong word for what’s happening, and I want to be exact about it, because the obvious objection is sitting right there: how is a country spending a trillion dollars on AI trying to slow it down? It isn’t. Washington is not slowing AI. It is slowing the open, fast, diffusive format of AI — the version that hands China a free copy — while sprinting as hard as it can on the parts that don’t diffuse. Build the data centers and the power plants flat-out, because physical infrastructure is a friction asset; nobody copies a gigawatt of baseload by reading a paper. Gate the models and the weights and the capability, because those are the things that leak. Accelerate what can’t be stolen. Throttle what can. That’s not a contradiction. That’s the whole doctrine in one line.

And it’s the national-security faction’s doctrine specifically — the same cluster I’ve called the Genesis faction, the procurement reformers and the NSC and the strategic-level players at Treasury and the War Department who treat the rentier economy as a liability. They want the domestic buildout fast and the global diffusion slow. The rentier class wants the opposite, because the rentier class makes its money selling the frontier at a premium. The Fable pull is the national-security faction winning a round.

Where the gate works, and where it’s late

I’ll be honest about the limits, because the strategy has one and it matters.

You cannot deny a copy of something to a man who’s already drawing his own. At the model layer, the gate is going up late. Two years ago, cutting China off from American frontier models would have starved its labs, because its labs were building off the American bleeding edge. That window is closing. China now ships its own frontier-grade open models — they no longer need the template, they’re writing it. Banning foreigners from Fable 5 doesn’t strand them; they’ll open a competitive Chinese model in another browser tab and keep working. At the level of the model itself, the barn door is half-open and the horse has a head start.

But notice exactly which models Washington pulled. Not the ordinary chatbot. It pulled Fable and Mythos — the models built on the cybersecurity-capable technology, the layer where a frontier capability can actually find and exploit the flaws in real systems. That’s the layer where diffusion still does strategic damage, where being six months ahead is the difference between your offense working and your adversary’s defense holding. The gate is being thrown up precisely where it still bites — the weaponizable edge — and left looser where the horse already left. That’s not sloppiness. That’s a regime picking its fights. And I expect more of that type of gatekeeping to be the norm, not the exception.

The sequencing

One last piece, and it’s the one that ties this to everything else I write about.

You do not want a thousand small uncontrolled shocks going off before your one big controlled one. A model leak, a capability surprise, a China-timed pop of the AI bubble — each of those is noise, and noise wrecks a precision operation. The larger move this whole project has been pointing at — the managed demolition, the rotation of capital out of the so called paper economy and into the real one, the high-energy sand thrown into the global machine at a moment of Washington’s choosing — that move needs a quiet board to land cleanly. Every uncontrolled disruption between now and then is a degradation of the eventual strike. Not that every variable can be controlled of course, but to the extent that the U.S. will attempt to control what it can..

The U.S. will continue to slow the ambient metabolism. You gate the leaks, you cap the surprises, you keep the frontier from spilling capability into the world on a schedule you don’t control. You preserve your optionality and your energy for the disruption you do control, the one you set off on your clock. Slowing the game down now isn’t timidity. It’s clearing the room before you do the thing you came to do.

That’s the whole argument. China wins fast, open, and cheap. America wins slow, gated, and dear. AI was born onto China’s side of the board, and Washington is dragging it, lever by lever and letter by letter, onto its own. The pulled model wasn’t an anomaly or an overreaction. It was a preview into the economic grind that is coming.

In healthcare and housing, we aren’t seeing “inflation” in the traditional sense of too much money chasing too few goods. We are seeing extraction. Concentrated interests have captured the necessary infrastructure of life, turning it into a captive market.

It creates the mindset of the vulgar extortionist. Because these industries function as cartels with little to no meaningful competition, they can set the price at the point of maximum pain and simply shrug off objections with a single question: “Where else are you going to go?”

Side promotion note: I’ll be working this beat for a new local substack called The Boulder Bulldog on the above themes, especially the flagrant fully legal scam that occurs when real estate political interests have no natural predators in its ecosystem.

But this post isn’t about necessities being used to extract from the many to the few, but in the more traditional U.S. equity market model of the moat concept, which is effectively a question about which sectors or stocks are best positioned to capture profit margins. Or to use a 1980s drug dealer analogy: why sell a kilo weed when you can sell cocaine at 2000% margin?

With that in mind… moving to the bleeding edge of the American technology stack:

Every American tech bubble of the last thirty years inflated in a sealed room — no foreign competitor to force prices down — and this is the first one where China is standing in the middle of it, commoditizing the product on a release schedule. Louis-Vincent Gave’s line was the frame: when China walks into a room, profits walk out.

This week, the casualty report. Three companies are the purest test of that thesis — OpenAI, Anthropic, and Google. The question every investor pouring money into them is really asking, whether they say it out loud or not, is this: are these the great profit-capturers of the AI age, the next Google-circa-2010 minting margins for a decade? Or are they about to be turned into utilities — kept alive, kept useful, and kept on a thin margin by a competitor that never stops building?

The honest answer is that none of the three escapes as a pure captor. But they are being commoditized into three different shapes, and only one of them has found something China can’t easily take… but the open question remains: What if being at the bleeding edge for 6 months doesn’t really matter in the grand scheme of things?

The thing being commoditized

Start with what “commoditization” actually means here, because it’s not one thing.

An AI lab sells across three layers. There’s inference — the raw cost of running the model, the price per token. There’s capability — how smart the model is, how far ahead of the free/cheap alternatives. And there’s deployment — how deeply the thing is wired into a customer’s actual work, the switching cost of ripping it out, i.e. the artificial ‘pain in the ass to change’ factor. This is a classic American tradition, long supported by both business and its state captured regulator mechanisms.

China is crushing the first two. Chinese open-weight models now run a few points off the global frontier and cost five to twenty-five times less per token. Inference is becoming electricity. Capability is following close behind. What China cannot crush by releasing a cheaper model is the third layer — the part where a model is so embedded in a company’s codebase and workflows that swapping it out means rebuilding the plumbing.

That third layer is the whole game: It’s the difference between selling a commodity and owning a customer. And the three labs sit in very different places relative to it.

OpenAI: priced like a captor, built like a utility

OpenAI is the most exposed of the three, and the gap between how it’s valued and what it earns is the widest in the industry.

The numbers, as of mid-2026: roughly $25 billion in annualized revenue, on a 33% gross margin. Hold on that margin. A healthy software business runs seventy to eighty percent gross margins — the cost of serving one more customer is almost nothing. OpenAI keeps thirty-three cents on the dollar, because the other sixty-seven go to the crushing cost of running the models. Inference cost hit $8.4 billion in 2025 and is projected at $14.1 billion in 2026. The company is on track to lose around $14 billion this year, burn roughly $27 billion, and doesn’t turn cash-flow positive until 2030.

Against that, the valuation: $852 billion at the last round, and in June it filed to go public at a number north of a trillion dollars.

That is a company priced as the great captor of the AI age while operating, at the margin line, like a utility that hasn’t found its rate base yet. ChatGPT has nine hundred million weekly users, which is a genuinely staggering distribution asset. But the core product — access to a frontier model — is precisely the layer China is driving toward the cost of electricity. OpenAI’s answer is to push into enterprise (now over forty percent of revenue), into coding with Codex, into ads, into anything with a stickier margin than raw model access. That push is the tell. It’s a company that knows the thing it’s most famous for is becoming a commodity, racing to attach itself to something that isn’t.

Anthropic: the one that found the moat

Anthropic is the same kind of company as OpenAI — a frontier lab, burning capital, racing the same Chinese price collapse. But it made a different bet, and the bet (at least for now) is working.

Anthropic’s annualized revenue hit roughly $47 billion in May 2026, and it did it by pointing almost entirely at business customers: about eighty percent of revenue comes from enterprises and developers, not consumers. Eight of the Fortune 10 are now customers. The standout is Claude Code, its coding product, which went from zero to $2.5 billion in annualized revenue in roughly a year, with more than half of that from enterprises.

Here’s why that matters in the language of the three layers. Coding isn’t a chat window you can swap for a cheaper one on a whim. Claude Code gets wired into a company’s repositories, its developer workflows, its build and review pipeline. Anthropic also authored something called the Model Context Protocol — a standard way to plug Claude into a company’s internal systems and data — and got the rest of the industry to adopt it. That’s the deployment layer, the one China’s cheaper models don’t automatically take, because the cost isn’t in the model, it’s in the integration.

One honest caveat, because the headline number flatters: Anthropic books its cloud-reseller revenue gross, counting the customer’s full spend as revenue and the cloud partner’s cut as an expense. That inflates the top line against peers who report it net. The $47 billion isn’t apples-to-apples with OpenAI’s $25 billion. But the structural point survives the asterisk: Anthropic sells the layer with switching costs, and that is the layer that holds when the model underneath becomes a commodity.

Google: the one that owns the cost stack

Google is in the strangest position of the three, and the most defensible — not because it captures the highest margin, but because it can survive the lowest one. This makes Google the China of the American frontier. Not by intent — by structure. It owns its stack, it doesn't need the margin, so it can sit at the commodity floor that drowns everyone forced to rent. China commoditizes the labs from outside the country. Google does it from the next desk over.

OpenAI and Anthropic rent their compute. Every token they serve runs on someone else’s chips — mostly NVIDIA’s, sold at NVIDIA’s roughly seventy-percent margin. Google designs its own. Its TPUs — custom AI chips it has built for a decade — let it run Gemini on hardware it owns end to end, stripping out the chip vendor’s markup that the other two pay on every token.

In a world where China is driving the price of inference toward the floor, the company that owns its own floor wins by default. Google can meet a commodity price that bankrupts a competitor renting GPUs, because its cost base is lower by the exact margin NVIDIA charges everyone else. And it doesn’t need the AI business to capture monopoly profits at all — it already has a search-advertising monopoly throwing off the cash. Google can afford to run Gemini as a near-utility, because being a utility on top of a vertically integrated cost stack, next to an existing cash monopoly, is a fine place to stand.

Put the three together and a pattern shows up. The lab selling raw intelligence is the most exposed. The lab selling integration has a moat. The lab that owns the hardware underneath everything can survive at any price. China sets the floor; the floor decides who’s left standing.

What this has to do with everything else

Here’s the part that connects to the larger argument I keep making, and I’ll keep it short.

The commoditization of these labs isn’t only a story about Chinese competition. It’s also a story about what Washington’s national-security faction wants. The same logic that has the U.S. government deliberately crushing NVIDIA’s chip margins — because a country can’t rebuild an industrial base while every input is priced for maximum extraction — applies to the intelligence layer too. Cheap inference is an input. The defense buildout, the scientific-research push, the autonomous-systems programs all need intelligence priced like electricity, not like a luxury good.

So China commoditizing the model layer from outside and Washington’s tolerance for that commoditization from inside point the same direction, the same way they did on chips. The labs financing this buildout — and the hyperscalers financing the labs — may be financing the compression of their own margins. That’s not an accident anyone planned. It’s the structure of the moment.

Then the state turned one off

While I was drafting this, another data wrinkle entered the chat.

On June 12 (last night, while I was in the middle of coding), three days after Anthropic released its newest top model to the public, the Commerce Department sent a letter — received, the company says, at 5:21 that evening — ordering it to cut off access to that model, Fable 5, and its more powerful sibling Mythos 5, for any foreign national anywhere on earth. The grounds were export controls and a national-security worry about a “jailbreak,” a trick for getting around the model’s safety limits. Anthropic complied within hours, then published a statement all but calling the reasoning a pretext: the jailbreak in question amounted to asking the model to read some code and fix its flaws, which it pointed out OpenAI’s GPT-5.5 does every day. Applied across the industry, the company warned, this standard would mean no new frontier model could ever ship. It was the first time in history a commercial AI model was pulled on a government’s order.

Look at the shape of the order. It doesn’t ban the model. It bans foreigners from using it — including Anthropic’s own foreign-national employees. Which raises the practical question the directive never answers: How is anyone supposed to enforce that? To actually verify the citizenship of every user, Anthropic would have to bolt a passport-checking bureaucracy onto a product used by hundreds of millions of people. And even then, any citizen could open an account and hand the password to a “foreign national” friend. The measure cannot stop the thing it claims to stop. The real adversary — a foreign state that wants the model — is exactly the actor with the means to proxy around a citizenship check. What the rule blocks is casual use. What it can’t block is the threat it names… in essence, a flimsy pretext.

When a security measure can’t achieve the security goal it cites, the stated goal usually isn’t the point. The point is the friction. Force Anthropic to either kill the model outright — which is what happened — or stand up an expensive, leaky verification regime it can never quite make work, and you have thrown sand in its gears: a cost imposed, a precedent set, a reminder issued. Washington has spent the last two years using exactly this kind of friction as a weapon at other people’s chokepoints. Now it has pointed it at one of its own labs.

The reminder is the part that matters for everything above. The question this whole piece has been circling — captor or utility — just got answered for Anthropic in real time, and not by the market. A captor sets its own terms. A utility runs at the government’s pleasure, switched on and off by people who don’t have to explain themselves. Washington just showed it can reach into a commercial AI company and take its best product dark between dinner and bedtime. Anthropic makes the best coding model in the world. It went dark anyway, because the state said so.

Why Anthropic, and why now — that’s where I’ll speculate, and flag it plainly as speculation. The pretext is thin enough to invite the question. Maybe it is nothing more than the state proving it can do this at all, and picking the most safety-vocal lab because that lab would comply quietly and set the cleanest precedent. Maybe it quietly favors whichever lab proves most useful to the national-security apparatus. I can’t tell you which, and the obvious story cuts both ways: Anthropic has been the industry’s loudest safety cooperator, red-teaming this very model alongside the government for thousands of hours before launch — which is either why it got hit, or beside the point. The narrow thing I’d say with confidence is this: the capability the state just revealed matters more than the case it used to reveal it.

But there's a deeper read here, and I'll give it its own piece in a few days, so take this as a promissory note. Pulling Fable might not be about Anthropic at all. It might be about the clock. China wins a fast, open world — copy-speed, open weights, instant diffusion. America wins a slow, gated one — energy, capital, chokepoints, control of the gates. If that's right, then the whole game the US is playing is to drag the contest off China's turf and onto its own: to convert a speed-game into a friction-game. And AI, which diffuses to everyone in six months, is the purest speed-game there is. Slowing it isn't fear of losing the race. It's a bet that the race itself is the format that favors Beijing — and that the winning move is to stop running and start gatekeeping. More on that soon.

Mostly utilities

On the open question of whether or not the frontier guys are going to be pushed into utility status, I think the answer is mostly going to be yes.

The old American playbook was to find a high-margin sector and wall it off — patents, lock-in, a captured regulatory state that serves as enabler — and this helps the dominant sector collect the rent for a generation. These three labs are the test of whether that playbook still runs against a competitor that commoditizes on a schedule and a government that can switch the product off between dinner and bedtime. It doesn’t. Most of this ends as utilities.

Not worthless — utilities. Metered, useful, enormous, and thin. OpenAI is the clearest case: priced like the captor of the century, earning thirty-three cents on the dollar at the exact layer China is driving toward electricity, racing into ads and enterprise because it can feel the model becoming a commodity under its feet. Google is the comfortable case — a utility by choice, because it owns the cost stack and has a cash monopoly next door. It doesn’t need the AI margin and won’t miss it.

Anthropic is the one with a shot at more, and even there I’d call it a slight moat, not a wall. Enterprise integration — code wired into a company’s pipelines, a protocol the rest of the industry adopted — is the one layer a cheaper Chinese model doesn’t automatically take. It’s real and it’s defensible. It is also not the seventy-percent rent the valuations assume, and the Fable pull just showed that whatever moat Anthropic digs, the state can fill back in with a single letter at 5:21 on a Friday.

So: utilities, with one thin moat among them. Intelligence itself — the core product, the thing a trillion dollars is being spent to make — becomes the cheapest input in the economy, and the profits, as Gave promised, walk out of the room.

The bigger question

There’s one more, and it might be the one that should worry the people writing the checks.

I had Fable 5 for about three days before the government took it back. I used it for real work, writing code. Is it better than what I had before? I think so. Probably, at the margin.

Sit with how thin that praise is, because the whole American AI economy is built on it being thick. The best model on earth, and the honest review is a shrug and a maybe or to borrow from some elementary intellectual property law jargon, potentially wouldn’t pass the PHOSITA test (Person Having Ordinary Skill in the Art). Meaning, the next iteration is not novel, but just the likely next step in an iterative progression of development.

The venture-capital model, the hundred-billion-dollar rounds, the talent wars, the trillion dollars of data-center capex — all of it rests on a single assumption: that the bleeding edge is decisively and durably valuable. That the best model is worth paying almost anything for, because the edge is where the returns live. This is how the American system has always worked. Fund the frontier, get there first, capture the rent before the pack catches up, then foreclose on competitors using a private-public “partnership” of corporate power backed by state regulatory authority.

But the edge is now both marginal and fleeting. Marginal, because “I think so, probably” is what the best model on earth buys you over the second-best. Fleeting, because in six months the edge diffuses outward — a cheaper Chinese open-weight model, or just the next release cycle, hands everyone the same capability at a twentieth of the price. So who pays the enormous premium to stand at the split-second frontier, when the frontier turns into a commodity by Christmas?

That doesn’t just squeeze three labs’ margins. It runs against the operating logic of the entire American machine, which has never once doubted that the edge is worth the chase. Venture capital, the IPO dream, the whole winner-take-all story rests on being first means getting filthy rich. That logic ran unbeaten for decades for one reason: nothing ever competed the edge away before the winners could bank it — no foreign rival fast enough, no discipline from inside. Both just arrived. The model isn’t wrong so much as long in the tooth, built for a game that had no other players and now run against a full field. Which leaves the question the American system has never once had to ask: what if the edge just doesn’t fuckin matter? “Almost as good, half a year later, nearly free” is starting to look like the winning hand — and if it is, the second mover was never behind. It was refusing to pay a fortune for a lead that melts before anyone can bank it: a few months at the frontier, bought at trillion-dollar prices, while the rest of the world arrives for free by Christmas.

The four largest American hyperscalers — Microsoft, Alphabet, Amazon, and Meta — plan to spend more than $700 billion building AI infrastructure in 2026. Add OpenAI’s Stargate commitments and the number clears a trillion dollars before most of the chips are even racked. All of it rests on one assumption: that the AI layer will eventually generate enough profit to justify the spend.

The consultants at Bain ran that math and found a hole. To support the capital being poured in, the AI business needs to throw off something like $2 trillion in annual revenue by 2030 — and comes up roughly $800 billion short.

That gap is what people mean when they ask whether AI is a bubble. The answer turns on one factor the financial press keeps underweighting.

Louis-Vincent Gave, who runs the research shop Gavekal and has watched Beijing flatten more industries than most Wall Street analysts have covered, says it in seven words: when China walks into a room, profits walk out.

The room used to be sealed

Every American asset bubble of the last thirty years inflated inside a sealed room.

The dot-com boom of the late 1990s had no foreign competitor at the frontier. When it burst, the survivors — Google chief among them — went on to earn extraordinary margins for two decades, because nothing on the outside ever forced those margins down. The platform giants that followed, Apple and Amazon and Meta and eventually NVIDIA, consolidated their sectors, bought or buried their rivals, and operated under antitrust regulators who had decided by the 2000s that enforcement was optional. Gross margins ran past seventy percent. NVIDIA’s reached 74.6.

A high-margin business with no competitor can tell a story about future profits more or less forever, because nothing outside the room forces those profits to either show up or die. That is the engine that inflated every recent American bubble. Cheap money helped. The sealed room mattered more.

This time the room is open

The current cycle is the first in a generation where the room is not sealed.

China runs one playbook, and it has run it in sector after sector: steel, aluminum, solar panels, shipbuilding, batteries, electric vehicles. Beijing floods the industry with cheap state-backed capital. Dozens of domestic firms pile in at once. Competition among them grinds margins to zero and then below it. The Chinese have a word for the resulting misery — involution, or neijuan, the internal grinding that compresses prices until a firm either drives its costs down or goes under. Chinese commentators use the term as a complaint. Washington should read it as the spec sheet for a weapon.

Beijing also refuses to let any single firm grow into the kind of politically untouchable champion that American tech became. When Ant Financial lined up the largest IPO in history in late 2020, the Chinese state killed it days before launch and spent the next two years disciplining Alibaba, Tencent, and Didi. Read in the West as Beijing strangling its own innovation, the move looks different from the production side: a government declining to mint its own trillion-dollar rentier. Every emerging sector gets pushed back down under a low-margin, highly competitive regime before it can capture the state the way American tech captured Washington.

In AI, the opening shot was DeepSeek. A Chinese lab trained a model rivaling the American frontier for under six million dollars, against the hundred million and up that comparable U.S. models cost to build. The American read at the time was that DeepSeek was a fluke, a one-off, a lucky punch that wouldn’t repeat.

It wasn’t a punch. It was a production line warming up.

DeepSeek was followed by Alibaba’s Qwen, then Moonshot’s Kimi, then Zhipu’s GLM, then MiniMax — a different Chinese lab shipping a frontier-grade open-weight model every few weeks. In one stretch this spring, four of them landed in twelve days. These are not toys. The best of them now run a few points off the very top of the global intelligence rankings, jostling with the frontier American labs, and they ship under open licenses that let anyone download them, run them, and resell them. The pricing is the part that should keep hyperscaler executives awake: frontier-tier Chinese inference runs five to twenty-five times cheaper than the American equivalent. A million tokens of a top U.S. model runs around fifteen dollars. The comparable Chinese model runs three to four, and with caching the input cost falls toward a few cents.

That is the room the $700 billion is being spent into. Not a single Chinese shock to absorb and move past, but a cadence — a new entrant every few weeks, each one dragging the floor price of intelligence lower and refusing to leave. The premise beneath the whole buildout, that frontier compute would stay scarce and expensive enough to protect fat margins for a decade, didn’t get disproven once. It is getting disproven on a schedule.

Why the profits walk out

The dot-com crash was a story about nerve. Companies with no earnings were priced as though earnings were coming, and one day the market stopped believing. The survivors then minted money for twenty years, because the room around them stayed closed.

The problem this time runs deeper than sentiment. The profits that justify the AI buildout may never arrive even if every investor keeps the faith, because there is now a permanent low-margin competitor in the room driving the price of the product toward the cost of making it. And not one competitor — a rotating bench of them, each Chinese lab undercutting the last. A company can build the most valuable infrastructure in history and still come up empty if the thing that infrastructure produces — intelligence sold by the token — gets commoditized to fractions of a cent by rivals that don’t need to earn a return on their capital to keep going.

That is the difference between a bubble that pops and a bubble that deflates. A pop is psychological, and it reverses. Structural margin compression does not reverse, because the competitor doesn’t leave the room.

The British made this exact mistake

The financial press reads Chinese deflation as Chinese weakness. Falling prices, the story goes, signal sick demand and a spiral Beijing can’t pull out of.

Britain read the United States the same way a century ago, and got it exactly wrong.

Through the long deflation of 1873 to 1896, cheap American manufactured goods flooded into a British market that had run the industrial world for a hundred years. British commentators called it “ruinous competition,” treated it as a kind of cheating, and assumed it would correct itself. It did not correct. It was the United States turning into the workshop of the world, and the falling prices were the discipline doing the building. Britain mistook the machinery of its own demotion for a passing weakness in the upstart.

The American commentariat is filing the same report on China today. The deflation that looks like frailty from inside a rentier economy is the competitive discipline that builds production superpowers. The people paid to explain markets on television can’t see it, because they sit downstream of the asset owners who profit from the opposite arrangement.

Where the money goes

So the profits walk out. The question for anyone holding the assets is where they walk to.

My working answer, and I’ll be straight that this is the part still unfolding, is that the capital rotates rather than evaporates. This is the core of what I’ve called the BlitzCorrection — the controlled demolition of the paper economy on top, with the proceeds flowing into the industrial buildout underneath. The same margin compression that hollows out the AI trade as an investment is exactly what the rest of the program needs to run. Cheap chips, cheap inference, and cheap autonomous systems are the raw inputs for the defense-industrial rebuild, the energy expansion, and the manufacturing base that Washington’s national-security wing is trying to stand up.

Here is the part that confuses people. Washington would, left alone, prefer the high margins. American power was perfectly comfortable with NVIDIA charging whatever NVIDIA wanted. But the national-security faction inside the U.S. government has decided those margins are now a liability, because a country can’t rebuild an industrial base while every input is priced for maximum extraction. So Washington has started crushing its own margins on purpose: Section 232 tariffs aimed at the chip layer, the January 2026 executive order barring stock buybacks at underperforming defense contractors, a widening effort to discipline a sector that thirty years of rentier policy priced into the stratosphere.

China, from outside the room, does the same thing to American margins that Washington is now trying to do from inside it. The two aren’t allies, and they aren’t coordinated. They point the same direction, and the AI sector sits at the spot where they meet.

Can you build a new room?

The capital expenditure is real, the chips are real, the infrastructure is genuinely getting built. Rotation explains where today’s money goes when the AI trade deflates. It doesn’t touch the harder question waiting underneath it.

For thirty years the American playbook was one move on repeat: find an emerging sector, let it consolidate, wall it off, collect the rent for a generation. The move worked because the room stayed sealed — no outside competitor ever arrived to force the price down. Patents, antitrust forbearance, platform lock-in all existed to keep it sealed long enough to extract.

That move now runs straight into Gave’s law. The moment a new American sector turns genuinely high-margin, it becomes worth China’s while to walk in. And once China is in the room, the margin that made the sector worth building is the first thing to go.

So here is the question every American planner is now facing, whether they say it aloud or not. Can the United States still protect a new high-margin sector? Or does every new sector have to crush its own margins on the way up, just to survive contact with a competitor that never stops building?

Two answers, pointing opposite ways.

One is the fortress. Seal the room again, this time by force — national-security designation, export controls, the trusted-partner clubs Washington is bolting together around chips and critical minerals. China doesn’t crush the margin in sovereign compute or defense electronics, because China is barred from those rooms by law. Build the wall high enough and the high-margin sector lives behind it. The rentier class can live with this, since the extraction simply continues under a flag.

The other is discipline. Any sector that touches the open world — exportable, consumer-facing, impossible to wrap in a security clearance — gets no wall, because you cannot wall off the whole planet. Those sectors meet China on China’s terms or they lose: thin margins, permanent competition, and an American capitalist class forced into the one thing it spent thirty years buying its way out of, which is actually competing, running lean, becoming the kind of production power the country last was in the 1890s.

Both cannot win, because they need opposite things. The fortress feeds the rentier; the discipline buries him. That contest is the factional war already running inside Washington — the national-security wing that needs cheap inputs to rebuild, against the rentier class that wants the wall and the rent behind it. Whichever side wins decides the answer.

I don’t know which way it breaks, and I’d distrust anyone who claims they do. What does look finished is the arrangement underneath the whole comfortable American century: build a thing, wall it off, coast on the margin. That part is ending, and it’s ending because of who is now standing in the room.

Which leaves the real question on the table, where it belongs. Can the United States build a single room China cannot walk into? And if the only way to survive is to crush its own margins before China does it first — will the people who run the country ever let it win that way?

The hyperscalers are doing two things at once. They are the apex of the rentier coalition that the BlitzCorrection is designed to reprice. They are also the primary financiers of the Genesis Mission infrastructure that absorbs the rotation when the repricing happens. That dual role is structurally unstable.

It works as long as Microsoft, Alphabet, Meta, and Amazon can fund their Genesis-direction capital expenditure out of monopoly cash flows that keep arriving faster than the new infrastructure has to demonstrate matching revenue. The moment the cash flow funding gets ahead of the revenue justification, the multiples that supported the apex valuation compress. That compression is the apex’s contribution to the rotation event.

As I’ve written in Old Money Becomes New Money, and Out with the old, the traditional historical rotation in both industrial and financial capitalism is the boom bust cycle—where the leading industrial sector of one era gets highly capitalized and accrues high profit margins—but then get repressed and subordinated by the next industrial technological emergence. The old effectively subsidizes the new.

So in the 1920s, U.S. Steel didn’t try to build a consumer automotive brand. They accepted their position as an upstream industrial input, leaving the consumer surplus, brand equity, and massive consumer-credit multiples to General Motors and Ford. In many ways, these dynamics shaped the capitalist factions in Roosevelt’s New Deal coalition.

But the hyperscaler paradox today is that the U.S. hyperscalers are attempting an unprecedented historical trick: they want to be the heavy steel mills and the high-margin consumer automotive fleets simultaneously.

1. The Strategy: Vertical Defense Against Subordination

The tech giants are acutely aware of the threat of subordination. They know that if they only provide raw compute and data centers, they will eventually be squeezed by downstream capitalist factions—the application developers, agentic networks, and enterprise workflows that actually interface with the end-user’s wallet.

To prevent themselves from being relegated to low-margin utility providers, they are executing an aggressive vertical defense:

• The Application Land Grab: Microsoft isn’t just renting out Azure servers; they are forcing Copilot into every enterprise seat. Google isn’t just selling TPU access; they are embedding Gemini directly into the core search monetization engine.

• Walled Garden Agents: They are trying to build proprietary agentic layers where the enterprise customer never leaves their software environment. If they control the digital agent that manages the corporate ledger or executes the commerce, they retain the premium software multiples.

2. The Break in the Strategy

Here is where their dual role breaks down under intense structural tension. To maintain the lead at the frontier “auto” (application) layer, they have to expand the “steel” (infrastructure) layer at a brute-force pace that is completely detached from near-term revenue.

This massive capital mobilization oversupplies the global market with raw computing power, driving down the marginal cost of intelligence.

When you overbuild infrastructure at that scale, you accidentally create the perfect conditions for your own subordination. Thanks to efficiency breakthroughs and model commoditization, the underlying “steel” of AI becomes incredibly cheap.

This allows lean, agile, third-party capitalist factions to build highly profitable, specialized applications without carrying a massive infrastructure debt load. The hyperscalers are effectively funding the very commodity curve that enables outside startups and enterprise players to bypass their high-priced proprietary applications.

Ultimately, they are fighting a war on two fronts: trying to build the asset-heavy industrial backbone of the next era while desperately clinging to the high-margin software valuations of the last one.

Video link: How the $750 Billion AI CapEx Boom is Reshaping Tech Industry Dynamics provides an analysis of how this massive spending cycle forces hyperscalers into vertical integration and pressures the enterprise software companies caught in their wake.

The $720 billion problem

U.S. hyperscaler capital expenditure for 2026 is guided at roughly $720-750 billion, a 67% year-over-year increase, locked into multi-year industrial deployment. That is not a normal CapEx cycle. It is a national mobilization happening through corporate balance sheets.

The question that follows is what revenue justifies that level of spending under standard return-on-invested-capital expectations.

Bain & Company published an analysis in 2024: To justify current and projected hyperscaler AI infrastructure spending at standard utility-scale return rates, the AI services layer needs to generate roughly $2 trillion in annual revenue by approximately 2030. That figure has been refined and pressured in the analyst community since publication, but the order of magnitude has not been disputed.

Current AI-attributable revenue across the apex is somewhere on the order of $50-100 billion annually. The gap between current run rate and target run rate is somewhere between 20x and 40x.

The gap closes through one of two paths. Either AI inference, agentic services, and AI-mediated commerce scale at rates without industrial precedent, generating the revenue ramp needed to justify the spend. Or the spend gets re-priced downward through multiple compression on the apex equities. Both paths are technically available. The market is currently pricing path one. Path two is the BlitzCorrection’s apex contribution—which is my base case.

The Hyperscaler Paradox creates the trap

The Hyperscaler Paradox is what makes this complicated. Microsoft, Alphabet, Meta, and Amazon are simultaneously the post-2008 monopoly winners and the primary architects of the Genesis Coalition’s physical infrastructure. They are not being displaced by the rotation. They are funding it.

That structural dual role would be sustainable indefinitely if the cash flow funding the CapEx came from segments insulated from the AI-revenue-ramp question. It does not. The cash flow funding hyperscaler CapEx comes from the same business segments whose forward earnings expectations depend on AI revenue scaling at the rates the spend assumes.

The apex is funding its own revenue ramp through CapEx that has to generate the revenue ramp to justify itself. That is a closed loop with a timing constraint. The closed loop works while the market trusts the assumption. The closed loop breaks when an external shock closes the gap. It might be a bad earnings report, or, as I’m starting to suspect, a Hormuz related supply chain shock bubble making its way through and blowing shit up (figuratively in this case).

Earnings as the test

Q2 2026 hyperscaler earnings, reporting in late July, are the first major data point.

Microsoft’s Azure AI growth rate is the cleanest single signal. Azure has been reporting AI services growing at triple-digit year-over-year rates from a small base. As the base grows large enough that the percentage growth rate has to compress for arithmetic reasons, market expectations remain priced for continued triple-digit growth. The first quarter that prints below that pace produces the first multiple compression event.

Meta’s CapEx-to-revenue ratio is the most exposed metric. Meta has been spending at ratios that historically only utility companies tolerate, on the assumption that AI-driven advertising efficiency and AI-mediated commerce will produce the revenue catchup. The first quarter where AI-attributable advertising revenue underperforms the spending ramp produces visible margin compression.

Google’s AI infrastructure utilization rate matters because Google has been deploying TPUs at a pace that depends on cloud customers absorbing the capacity. If utilization rates stall while CapEx continues, Google’s cost-per-revenue-dollar in the cloud segment deteriorates visibly.

Amazon’s AWS AI revenue trajectory is the most opaque because AWS doesn’t break out AI services as cleanly. The signal will come through cost discipline. If AWS margins compress while CapEx grows, the market reads it as the gap.

Q3 2026 (late October) is the confirmation cycle. Q4 2026 (late January 2027) is the make-or-break window. By the time Q4 prints, the market has had two consecutive quarters to assess whether the revenue ramp is keeping pace with the spend.

If both Q2 and Q3 confirm the ramp, the apex pivot continues to work without market disruption. If either misses meaningfully, multiple compression at the apex begins independently of any geopolitical trigger. By Q4, the pattern is locked in either direction.

But again, earnings is just one pathway here, and i’m not ruling out other external forces—caused by the friction forced into the global economic order by the U.S. attack on Iran and subsequent hormuz closure—and of course the ever increasing possibility of further kinetic action stacking even more economic friction into the system.

The DeepSeek complication

Another variable American industrial planners at the Genesis level must contend with is that the U.S. doesn’t stand alone in terms of sector profit margins being controllable at the top of the American elite. Foreign competition—particularly from China introduces margin crushing MUCH earlier in each technological development wave. I should probably do a deeper dive into the unique ways that China manages its market economy, but for now i’ll just quote Gavekal’s Louis Vincent Gave’s line about Chinese competition: “When China walks into the room, profits walk out”.

Chinese AI commoditization changes the calculation in ways the apex cannot ignore.

DeepSeek published its V3 model in late 2024 with a reported $5.6 million training cost. The reported figure was the marginal cost of the final training run, not the total CapEx. True CapEx for DeepSeek’s parent organization is somewhere between $1.3 and $1.6 billion. That is still an order of magnitude below U.S. hyperscaler scale, and the gap continues to widen as U.S. CapEx accelerates.

The strategic implication is what matters. Chinese model performance has been closing the gap with frontier U.S. models on standard benchmarks at a fraction of the spending. NIST CAISI’s September 2025 evaluation of DeepSeek’s earlier models found U.S. systems still leading by twenty percent or more on demanding tasks. The April 2026 evaluation of DeepSeek V4 Pro measured the strategic-application gap at approximately eight months. The Stanford AI Index 2026 reported a 2.7% Arena gap on the same comparison window.

Eight months and 2.7%. That is a closing gap. It is not parity, but it is rapid enough that the U.S. hyperscalers cannot reduce CapEx without ceding the frontier. They have to spend faster to maintain the lead. The faster they spend, the larger the revenue requirement gets. The larger the revenue requirement, the harder it becomes to satisfy through commercial AI services that DeepSeek-class models can substitute for at lower cost.

China’s commoditization strategy does not need to win the frontier race. It needs only to compress the time the U.S. apex has to convert frontier CapEx into apex-justifying revenue. Every quarter that gap closes, the U.S. apex’s path-one option (revenue ramp) becomes harder to execute and the path-two option (multiple compression) becomes the residual.

What apex compression looks like

The compression scenario is mechanical, not cataclysmic.

Microsoft trades at roughly 35x forward earnings. Meta at roughly 28x. Alphabet at roughly 25x. Amazon at variable multiples depending on segment treatment but broadly in the same range. These are premium multiples justified by AI-revenue ramp assumptions.

If the ramp underperforms, those multiples compress toward utility-infrastructure benchmarks. Microsoft to 22x. Meta to 18x. Alphabet to 18x. Amazon to a similar range. That compression alone, without any movement in earnings themselves, produces 25-40% drawdowns in the most heavily weighted index components. Combined with smaller compression in second-tier names, the S&P 500 composite drops 20-30% in the months following the trigger quarter.

That is the BlitzCorrection at the apex level. A multiple repricing event driven by an earnings cycle or external shock that exposes the math the previous multiples were assuming away. Potentially smaller and more contained than 1929 or Lehman, structurally different from both.

The receiving basins absorb the released capital. The Genesis Coalition infrastructure that the apex was funding continues to operate, because the operational CapEx is already deployed. The apex compression releases capital from passive index ownership of the old composition into passive index ownership of the new composition. That is the rotation.

The Hyperscaler Paradox in tension

The apex theoretically has an out. Microsoft could explicitly reposition itself as a Genesis Coalition industrial financier, accept lower software-business multiples, and use the cash flow to fund infrastructure that the next decade of AI deployment needs. Alphabet could do the same. Meta has been signaling something like this approach already through its public CapEx guidance.

The market punishes that repositioning even when management explains it. The same passive-bid mechanism that supports apex valuations also constrains the apex’s ability to voluntarily accept lower multiples. Index funds buy the apex by weight. When management announces that future returns will be utility-grade rather than software-grade, the multiple compresses anyway. The apex cannot quietly transition. The transition is visible by definition because the transition shows up in earnings releases.

The Hyperscaler Paradox in its full version is this: the apex can pivot to Genesis Coalition financier role, but the pivot itself triggers the multiple compression that the BlitzCorrection produces. I suspect the architects and industrial planners know this. The pivot is happening anyway because the alternative is the apex actually losing the AI race. The compression is built into the path either way.

The apex does not get to choose between continuing as the rentier coalition’s apex and becoming the Genesis Coalition’s primary financier. The apex has already chosen the second. The market has not yet repriced for it.

When the math comes due

The earnings calendar imposes a forecast structure on the BlitzCorrection that complements the Iran physical-markets variable. The trigger condition is positioned to fire on either path.

If Iran physical disruption flips the algorithmic narrative first, apex compression follows quickly because recession-pricing models cut AI revenue forecasts while the ramp is already underperforming. If apex earnings disappoint first, Iran disruption becomes the legitimizing narrative for the compression that was already happening through earnings discipline. The two variables are coupled. Either can fire alone. Both firing simultaneously is the most likely path because the conditions reinforce each other.

The strategic planning core

I’ve been working on this book project for awhile, and building out a map of distinctions and factions is a task i’ve been putting off, in part because it’s tough to define in a fluid environment—is subject to change as things evolve. I do have questions about some potential areas where margin crushing might still be in order, but that, like anything involving the margin crusher, might be more of a timing question than anything else.

The Genesis Architecture is what i’ve tentatively titled the book, but I pulled the idea from one the Dept of Energy (DOE) initiative called the The Genesis Mission:

The Genesis Mission unites DOE National Labs, industry, academia, and more to harness AI for breakthroughs in energy dominance, discovery science, and national security.

In essence, tying in multiple sectors and eventual factions that will exist in the Genesis Coalition or Architecture. This is an EMERGING phenomenon, which naturally makes nailing it down with precision much harder in advance. With that said, i’ll lay out a long ass primer on how I view it coming together.

The Genesis Architecture is my name for a strategy the U.S. government is executing in plain sight: the deliberate retirement of a digital-speculative, petrodollar economy and its replacement with a physical-industrial one built on a sovereign monopoly over energy and machine intelligence. The term is lifted from above mentioned actual federal program — i.e. Genesis Mission, which fuses the national labs, industry, and academia to aim artificial intelligence at “energy dominance, discovery science, and national security.”

Four moving parts: The Architecture is the whole plan. The Genesis Mission is its engine — the industrial and scientific sectors being stood up. The Natty Greenback is the fuel — U.S. LNG and Henry Hub pricing closed into a sovereign energy loop that supersedes the petrodollar. The BlitzCorrection is the trigger — a managed demolition of the paper economy that rotates capital out of the old order and into the new.

Underneath sit two coalitions. The Genesis Coalition — sovereign compute, frontier energy, defense reindustrialization, domestic chips and critical minerals, regulated stablecoins, the bioeconomy — is being capitalized and shielded. The Rentier Coalition — the passive-index bid, the platform and pharmaceutical toll-collectors, the cost-plus defense primes, the housing-finance and consumer-debt complex — is being compressed by tariff, statute, and procurement. The hyperscalers straddle both: the apex monopolies of the old order, financing the physical build of the new.

None of it runs on its own. Named institutions and named people execute it — Treasury, the NSC, the Fed, OSTP, the Pentagon — through executive orders, sovereign procurement, and export controls, each with a signature and a docket. And it is still emerging: a phenomenon in motion, easier to name than to pin down, and subject to revision as the factions harden.

Naming and citing the architecture

The National Security Council, restructured under Trump’s National Security Presidential Memorandum 1 (shorthand NSPM1), with the technology directorates moved above the traditional diplomatic and regional desks.

The NSC is now an industrial-policy machine wearing a foreign-policy hat. Quiet change 1, hiding underneath the Trumpian noise that’s distracted so many.

Treasury under Secretary Scott Bessent runs sovereign-investment programs and the sanctions architecture through the Office of Foreign Assets Control. OFAC‘s May 1 sweep against Iran under Executive Order 13902 was a Treasury operation, with named signatories and a public docket. It was not “the architecture acting.” It was a sub-agency of Treasury cutting off Iranian access to dollar settlement.

Inside the Federal Reserve, the Bessent-Warsh dynamic is now operational. Kevin Warsh, sworn in as Fed governor in May, gives Treasury a working channel into Fed policy that didn’t exist while Powell held the chair alone. The operational precedent is Steven Mnuchin’s November 2020 move, when Treasury unilaterally ordered the Fed to wind down the CARES Act emergency-lending facilities — Treasury override of stated Fed preference is a six-year-old template rather than a new invention. A major thing i’m watching, along the lines of the big 1979 moves that previously restructured the U.S. global political economy i.e. the Carter appointment of Paul Volcker to the fed, which again PREDATED the Reaganite shift that takes the public relations narrative credit, is the dynamic between the fed and treasury. The fed in 1982 didn’t operate like the fed in 1972. And the fed in 2028 might also be a subordinated force UNDERNEATH the treasury dept. Something to keep an eye on.

The White House Office of Science and Technology Policy (OSTP) and the President’s Council of Advisors on Science and Technology (PCAST) are the directorates channeling federal research funding toward sovereign-aligned compute infrastructure: the Stargate-adjacent buildout, the national supercomputing platforms, the AI-for-science programs. OSTP coordinates the executive branch’s R&D portfolio; PCAST is the outside-advisor body that signs off on strategic technology priorities.

The Department of Government Efficiency (DOGE), operating under Executive Order 14210, removes civil-service personnel whose institutional reflexes slow architectural decisions — the objective is clearing veto points rather than reducing headcount. In other words, DOGE and Elon take the bad PR. But the quiet part is effectively improving vertical network speed, removing the normal civil service friction.

The Department of Defense‘s acquisition reform faction operates through two instruments.

1. The SPEED Act expanded the dollar thresholds at which the Pentagon can buy commercial items rather than custom defense systems, which is how Anduril and Palantir win contracts that previously would have gone to Lockheed or Raytheon.

2. Executive Order 14372 prohibits stock buybacks and dividends by defense contractors whose major programs are underperforming, which is a direct sovereign ceiling on the cost-plus rentier model.

These institutions don’t negotiate with Congress to execute. Consensus is in hand at the level that matters, and decisions deploy through executive orders, sovereign-procurement vehicles, open-ended federal contracts, export controls, and sovereign settlement rails — each with a public number, a named signatory, and a docket on file.

The Genesis Coalition (capital side, emerging and ascendant)

Six clusters.

1. Defense-Industrial 2.0. Anduril (Brian Schimpf, Palmer Luckey), Palantir (Alex Karp), SpaceX (Elon Musk), Shield AI, Saronic, Epirus. These firms won allocations under SHIELD, the Pentagon’s $151 billion multi-vendor contracting vehicle for unmanned and AI-driven defense systems, announced March 3, 2026. They posted unit-cost results at Gauntlet I, the Pentagon’s first competitive procurement run for low-cost attack drones, held at Fort Benning in March 2026.

2. Sovereign-aligned compute via the Stargate consortium. OpenAI (Sam Altman), Oracle (Larry Ellison), SoftBank (Masayoshi Son), the UAE sovereign tech vehicle MGX (chaired by Khaled bin Mohamed bin Zayed), with CoreWeave and Crusoe at the infrastructure edge. A $500 billion, 10-gigawatt joint venture with disclosed capital commitments and a governance structure on file with the SEC and the relevant foreign-investment review bodies.

   1. Additional note: watch how the U.S. bonds its vassals to the compute stack—the MGX stake up above is the opening move, and it spreads from there. UAE first, then Japan, eventually Europe. And “bond” is pulling double duty on purpose—it’s the capital you post AND the leash, like a bail bond. You’re out walking around, sure, but you’re tied to the post, and the post is your own sovereign wealth. Their money builds the racks they don’t control. Call it a consolation prize if you’re feeling generous. But it’s closer to a protection racket—the roughing-up IS the pitch. Letting the GCC guys get whacked around by Iran, threatening to grab Greenland, and the Donald romping through the alliance system like a wrecking ball—all of it proves the allies can’t secure themselves OR stand up their own stack. The shakedown and the sale are the same move…. Nice alliance you got there, shame if something happened to your strait—but here, buy into our compute. They aren’t getting cut out completely. They get first dibs in the extended Pax Americana, a.k.a. Pax Silica—same tribute racket as the petrodollar, just priced in compute instead of crude.

3. Frontier energy. On the natural gas side: Cheniere, Venture Global, Sempra Infrastructure. On the advanced nuclear side: TerraPower (chaired by Bill Gates), NuScale, X-energy, Oklo, Fervo. Each one is a public or near-public firm with a CEO, a balance sheet, and either a 20-year LNG sales agreement indexed to the Henry Hub benchmark price or a Nuclear Regulatory Commission license application in process.

4. Industrial reshoring. TSMC Arizona, Intel Foundry, Samsung Texas, Micron New York on the semiconductor side. MP Materials and Lynas USA on the critical-minerals processing side. Plants in physical U.S. locations, run by named executives, financed through specific CHIPS Act tranches authorized by Commerce Secretary Howard Lutnick.

5. GENIUS-Act stablecoin issuers. Coinbase (Brian Armstrong), Circle (Jeremy Allaire), Tether (Paolo Ardoino), PayPal’s PYUSD. Their reserves are disclosed quarterly under the GENIUS Act of 2025 — the stablecoin regulation that requires issuers to back every digital dollar with short-duration U.S. Treasury securities — and their Treasury holdings get reported monthly.

6. Bio-industrial. BioMADE, Ginkgo Bioworks (Jason Kelly), Twist Bioscience, Recursion. These firms operate inside BIOSECURE Act compliance frameworks (the statute that restricts U.S. federal contracting with Chinese-linked biotech firms, enacted as a title of the FY2026 National Defense Authorization Act (Public Law 119-60, signed December 18, 2025) rather than as a standalone law), with active contracts from DARPA (the Defense Advanced Research Projects Agency, the Pentagon’s blue-sky R&D arm) and pipelines through BARDA (the Biomedical Advanced Research and Development Authority, the HHS office that funds medical countermeasures).

The Rentier Coalition (capital side, being compressed)

Compression is being executed against specific firms:

The passive-bid complex: BlackRock (Larry Fink), Vanguard, State Street, with roughly $30 trillion in combined assets under management. The legacy chip and platform rentiers: NVIDIA (Jensen Huang), whose 74.6% gross margin is being compressed toward defense-prime utility values under Section 232 of the Trade Expansion Act of 1962 (the national-security tariff authority Trump used in his January 2026 proclamation on advanced semiconductors). Apple’s App Store toll. Meta and Google’s ad monopolies. The legacy defense primes: Lockheed Martin (Jim Taiclet), Northrop Grumman (Kathy Warden), RTX, Boeing Defense, General Dynamics, whose cost-plus contracting economics the Gauntlet leaderboard model is positioned to dismantle. The pharmaceutical manufacturer and pharmacy benefit manager (PBM) oligopoly: Pfizer (Albert Bourla), Merck, Johnson & Johnson, AbbVie on the drug-manufacturer side; CVS Caremark, Express Scripts, OptumRx on the PBM side, where the benefit managers pocket the spread between what insurers pay them for a drug and what they pay the pharmacy that actually dispenses it; UnitedHealth and Elevance on the insurer side. The three layers are nominally separate businesses but functionally one power bloc. CVS owns Aetna. Express Scripts is inside Cigna. OptumRx is inside UnitedHealth. The Visa-Mastercard duopoly. The housing-finance bloc: Blackstone, Invitation Homes, Lennar, D.R. Horton, and the mortgage-backed-securities issuers. The student-loan complex: Sallie Mae, Navient.

The margin compression at this layer means these firms, by name, losing pricing power against specific statutes signed by specific officials… again, borrowing from American history. Railroads ascend—then become the utility for emerging sectors. Steel ascends, then margins crushed and become utility for the emerging automobile sector. The shift here is that nobody is pretending the magic of the market is going to produce this. It’s a quiet return to industrial planning, hidden beneath the Trumpian chaos.

The Hyperscaler Paradox

Microsoft (Satya Nadella), Alphabet (Sundar Pichai), Meta (Mark Zuckerberg), Amazon (Andy Jassy) are doing two things at once. They’re the apex post-2008 monopoly winners. They’re also the primary financiers of the Genesis Coalition physical infrastructure being built right now. Microsoft funding OpenAI inside Stargate. Amazon funding sovereign-cloud architecture. Meta funding open-weights AI releases that subordinate competing API margins. Alphabet funding expansion of its Tensor Processing Units (Google’s in-house AI chips, designed as alternatives to NVIDIA’s GPUs). Hyperscaler capital expenditure for 2026 runs $720 to $750 billion, sixty-seven percent above 2025. The pivot happens inside CapEx committees, signed by CEOs, ratified by boards. Old money becoming new money is a series of capital allocation decisions with quarterly accountability.

The Genesis Architecture is the sum of these institutions’ decisions and these firms’ capital flows. Bessent signs Treasury directives, Lutnick signs Commerce orders, Trump signs executive orders. Karp signs Palantir’s earnings statements; Fink, together with his counterparts at Vanguard and State Street, presides over the passive bid; Nadella approves the Stargate transfers that fund OpenAI’s compute.

The passive bid deserves a fuller description, because it is the most consequential and least understood piece of the architecture… and obviously something i’ve tried to hammer and repeat on this substack, i.e. Schrödinger’s ceasefire, Old Money becomes new money etc. The talking of markets up and down is creating exit liquidity for massive funds to sell into.

The three firms — BlackRock, Vanguard, State Street — collectively run roughly $30 trillion in indexed assets. Their funds buy whatever stocks the benchmarks contain at whatever price the market is showing, mechanically, regardless of valuation. Mike Green’s work has been invaluable to me here, watch his interview with Maggie Lake for long form discussion on it.

The flow comes from 401(k) auto-deductions and target-date funds, roughly $40 billion a month into VOO and SPY alone. When rentier insiders sell their concentrated positions — Palantir’s Peter Thiel cashing out roughly $300 million in shares, hyperscaler executives exercising options through 10b5-1 plans, defense-prime CEOs collecting bonuses denominated in stock that they then turn into cash — the buyers on the other side of those trades are the passive funds. The funds absorb the supply without negotiating price. The retirement-savings stream that built the rentier valuations over the post-2008 period is now funding the orderly exit out of them, and the eventual repricing falls on the retirement accounts holding the index funds. Whether or not this will feel and look like the massive looting the occurred in Russia in the 90s will largely depend on how quickly the new coalition can soak up weight in the S&P index, to avoid the massive popular discontent that will happen when the mass of the American public loses their ass in a financial crash, i.e. how fast the financial and economic recovery can occur. (note: 1 major difference is that the vast majority of Americans own jack shit in terms of assets, so the hit here will come more in the form of potential layoffs/unemployment, etc).

Bottom line: All of this is documented in specific filings, attributable to specific people. It all runs THROUGH trump 1-biden and trump 2. Not a specific administration factional alliance limited to Trump 2.0.

The Trump administration has now fired nine identical Iran-deal head-fakes since the Hormuz closure began in February. Each one follows the same script. A tweet, a Rubio readout, a Witkoff envoy gesture, an Axios scoop about a 60-day MOU “largely negotiated.” Iran issues a denial within hours. The headline gets walked back inside a day. And yet WTI drops every single time. Twenty dollars at the peak. Lately just one.

Twitter is perplexed. Who is still being fooled by this? The same hedge funds that got burned on head-fakes one through eight? The same energy analysts who already wrote the “this won’t actually happen” Substack post six weeks ago? Surely after the third repetition of the same trick, the market participants who actually move money should have stopped responding.

They have stopped responding. They are not the audience.

The audience is the algorithms.

How the trick actually works

Three layers of automated machinery sit between a Trump tweet and a WTI price drop. None of them require a human to read the tweet, weigh whether the claim is true, or form a view about whether a deal is actually coming. They respond to the existence of the headline itself.

Layer one: high-frequency trading market makers. Firms like Citadel Securities, Jane Street, Virtu, and a handful of others supply most of the moment-to-moment buy and sell quotes in oil futures during normal trading hours. Their algorithms read news feeds in milliseconds. When a headline fires that classifies as “bearish for oil” — anything that sounds like Iran-deal progress, Hormuz reopening, peace negotiations — the algorithms pull their bids and widen their spreads to protect themselves from sudden moves. The available liquidity at any given price level immediately thins out, and the next sellers find a market that’s stepped back.

Layer two: the passive bid through commodity and energy ETFs. An ETF, or exchange-traded fund, is a basket you can buy a single share of — one share gives you a slice of everything inside it. BlackRock, Vanguard, State Street, and the other major asset managers run a small constellation of energy-tracking funds: USO and BNO hold oil futures directly, XLE and XOP hold energy stocks. Ordinary investors hold these as their “energy exposure” without thinking about it much.

Here’s the part that matters. When you sell your shares of an oil ETF, the fund has to come up with cash to pay you, and to get that cash it sells some of the actual oil futures it’s holding. That forced sale of the underlying holdings is called a redemption. One person selling their ETF shares is nothing. But when sentiment around oil turns negative and thousands of investors and automated portfolio systems all trim their “energy” exposure at once, the funds have to dump a large volume of real oil futures into the market to meet all those redemptions at the same time. The price drops. Not one of those sellers was making a judgment about Iran. They were trimming a line item in a diversified portfolio because “energy looks weak this week,” and the fund’s mechanical selling moved the actual oil market on their behalf.

Layer three: options dealer gamma. This one needs unpacking, because it’s the least understood and the most powerful.

Start with who the “dealers” are. A dealer, also called a market maker, is a firm that stands in the middle of trades. When you want to buy an oil option, the dealer sells you one; when you want to sell, the dealer buys it from you. The dealer isn’t betting on which way oil goes. It makes its money on the spread — the small gap between its buy price and its sell price — and it wants to stay neutral, not exposed to the market moving in either direction.

An option, in turn, is a contract that gives someone the right to buy or sell oil at a set price by a set date. Traders use them as bets or as insurance.

Here’s the dealer’s problem. After selling a pile of options to clients, the dealer is left holding exposure to the oil price moving. To cancel that exposure and get back to neutral, it buys or sells actual oil futures. This is called hedging. But the amount of hedging it needs isn’t fixed — it changes as the price moves. “Gamma” is just the word for how fast that hedging requirement changes.

When dealers are “short gamma,” which is the typical position after they’ve sold a lot of options, the math forces them to trade in the same direction the market is already going. Price falls, they’re forced to sell more to stay neutral. Price rises, they’re forced to buy more. Their required hedging amplifies whatever move is already underway instead of cushioning it.

Options cluster at round-number prices. For WTI right now there’s a heavy concentration around $95-100. That’s where dealer exposure is largest, so that’s where this amplification force is strongest. When a bearish headline pushes oil down, the short-gamma dealers near that concentration are mechanically forced to sell more oil futures to keep their books balanced, which pushes the price down further. They aren’t trying to suppress anything. The math of staying neutral requires them to sell into a falling market, and the falling market feeds on itself.

What a Trump tweet actually triggers

The chain runs in milliseconds, not minutes.

Trump tweets about Iran deal progress at 8:34 AM Eastern. By 8:34:00.4 the HFT algorithms have classified the headline, pulled bids, and widened spreads. Within the next minute, sentiment-system feeds at the major brokerages and ETF managers mark the energy sector “weakening.” Over the next ten minutes, retail order flow and pre-programmed institutional rebalancing kick in. ETF redemptions trigger underlying futures sales. Dealer gamma positioning at the upper strike concentration unwinds — dealers who were short futures to hedge against the price reaching the strike now need fewer of those hedges, and the way these positions unwind often amplifies the down-move further. By 9:00 AM, WTI is down four dollars and Twitter is starting to react.

No human, in that entire sequence, has read the tweet, weighed whether the claim was credible, or formed a fundamental view about whether a deal is actually coming. The whole thing happened on autopilot.

Iran will issue its denial a few hours later. It won’t reverse the move. The down-move from the original headline stays in place, the Trump administration gets its suppression, and no human had to be deceived. Why the denial doesn’t land is the most important part of the whole mechanism, and it gets its own section.

The feed only points one way

Here’s the question lurking under all of this: if the algorithms respond to headlines, why doesn’t Iran’s denial move the price back? Iran denies the deal progress every single time. The denial is real news. Why doesn’t it register?

Because the algorithms aren’t monitoring a neutral global news feed. They’re weighted heavily toward fast, English-language, high-trust, high-velocity channels — and the single highest-velocity political-headline channel on earth is Donald Trump’s social-media account. When Trump posts that an Iran deal is close, it hits every wire service, every brokerage sentiment system, and every trading algorithm within seconds, tagged as a top-priority signal from the head of the U.S. government.

When Iran denies it, the trading algorithms barely register the denial — and not because it’s unavailable or untranslated. The denial goes out through the foreign ministry, through Press TV and IRNA, and through a dense ecosystem of English-language Middle East OSINT channels on Telegram. Middle East Spectator, MaxOsint, Tabz, and dozens of others, many with six-figure followings and a faster, more accurate read on the actual situation than the wire desks, carry it in English within minutes. Any human following the conflict sees it in real time.

The machines don’t. The sentiment systems that move price are tuned to the wire services — Bloomberg, Reuters, AP — and to a short list of verified, high-velocity, high-follower accounts. Trump’s account is integrated into every one of those systems as a top-priority signal. A Telegram OSINT channel that called the last six head-fakes correctly is not in the feed at all. The denial exists in the world, in English, accessible to anyone who looks. It does not exist in the feed the machines are reading.

That asymmetry is the whole reason the suppression works on repeat. Trump can move oil down with a sentence. Iran cannot move it back up with a sentence, because its sentences aren’t in the machine’s input set. The only thing Iran can do that reliably registers is kinetic — a tanker attack, a missile launch, a strike on infrastructure. Those events show up directly in price and volatility data, bypass the language and channel filters entirely, and force the algorithms to react. Words versus violence, and only one side’s words are in the feed.

That’s why the pattern is so stable. The U.S. side gets to play the game with text, on the cheap, as often as it wants. The Iranian side can only answer with action, which is expensive, escalatory, and rare. The game is rigged at the level of which headlines the machines even hear.

And notice the irony, because it cuts against the reflexive assumption that Iran is the unreliable one. Of the two actors, Iran has been the consistent one. It has denied deal progress identically, every time, for nine straight head-fakes. The unreliable claims are all coming from the U.S. side — nine assertions of an imminent agreement, nine failures to produce one. A system that weighted accuracy or track record would have learned to discount Washington and take the Iranian denials at face value by now. But the feed doesn’t weight accuracy. It weights velocity, language, and channel. The least reliable actor, equipped with the fastest English-language megaphone, wins every round.

Trump didn’t invent this. But his administration has industrialized it, and the technique is spreading. Once you understand that a well-timed, high-velocity, English-language social post can move a global commodity market that a foreign government’s official statement cannot, you have a tool. More actors are figuring it out. The feed asymmetry is becoming a standard instrument of market and information operations, and it will outlast this particular Iran cycle.

Why the trick is wearing out

The reason this works is that the algorithms are trained on patterns. The reason it’s stopping working is that the algorithms are starting to recognize the pattern.

Head-fake one in March produced an $8 drop with eight sessions of mean reversion. Head-fake three on April 17 produced a $21 drop in two hours, then snapped back violently inside four sessions. Head-fakes four through eight produced $4-7 drops each, with two to six sessions of recovery. Head-fake nine, today’s Axios MOU scoop with Witkoff and Rubio working it through Pakistani mediators, produced a one-dollar drop that faded inside a single session.

The mechanism is intact. What’s eroded is its effectiveness. The HFT classifiers are weighting “Iran deal progress” headlines less aggressively each time. The ETF sentiment systems are absorbing the noise as background rather than signal. The dealer gamma positioning is starting to adjust on the upside, anticipating the eventual price reset rather than purely defending the suppression.

This is a real-time stress test of the apparatus. The Trump administration has been firing the only tool it has, on the same target, on increasingly short intervals. The tool is degrading. By the time the underlying physical reality (the Strategic Petroleum Reserve approaching its statutory floor in late August, commercial inventories slowly bleeding, the absence of any actual Iran deal) reasserts itself, the suppression machinery may already be too worn out to dampen the price move that follows.

So who’s actually being fooled?

Some humans, sure. There are retail traders short oil right now on the genuine belief that a deal is coming, and they’ll get carried out when it doesn’t. There always are. But they are not who the operation is aimed at, and they are not what moves the price. The “who’s the sucker” framing assumes human credulity is the mechanism. It mostly isn’t. The mechanism is automated market plumbing — high-frequency quotes pulling back, ETF redemptions forcing sales, dealer hedging amplifying the move — that fires on textual inputs, executes its programming, and moves the price before any human has finished reading the headline. The target audience is the machines. The humans who get fooled are collateral.

The Trump administration figured out, somewhere in the early weeks of the Hormuz closure, that this plumbing was an available tool. They have been using it accordingly. The Iran denials don’t matter because they were never the point. The headlines were never primarily trying to convince humans of anything. They were inputs into the algorithms that move price.

When the algorithms finally stop responding — when the diminishing-returns curve hits zero — the tool breaks. That moment is approaching faster than it was a month ago.

A Bloomberg-syndicated piece on OilPrice.com argues that China’s potential return to aggressive crude buying could be the next global price shock. The article frames China’s current demand restraint as soft power — Beijing holding back so its Asian trading partners (Japan, South Korea, India, the broader ASEAN importing bloc) don’t get crushed by high oil prices. That part of the framing is correct. China is using muted import behavior as a regional-leadership signal, and Asian allies are benefiting from the restraint. What the article misses is that the soft-power play is being neutralized in real time by the U.S. suppression apparatus, and the snap that will eventually expose it is engineered for a window when China can’t shield those allies. China isn’t the shock-maker. China is the body inside the pincer, and the pincer is closing.

How the suppression apparatus actually works

Bessent at Treasury has been running the Strategic Petroleum Reserve down at a 1.4 mb/d clip — the SPR sat at 411 million barrels pre-crisis and is now under 375 million per this week’s EIA Weekly Petroleum Status Report for the week ending May 15. The IEA-coordinated international release has delivered roughly half of its 400-million-barrel commitment.

The SPR-dumping channel is operating at full throat. So is the talk-down channel: every administration statement since February has been calibrated to suppress, never to confirm, the physical reality. The passive-bid asset complex — BlackRock, Vanguard, State Street commodity ETF flows — responds mechanically to the sentiment-tagged signals these statements generate. HFT options dealers are net short gamma at the $95-100 strike concentration on WTI, which means as price approaches the strike, dealers must sell futures to remain delta-neutral. The mechanical price ceiling does the suppression work without any human intervention.

Four channels. All operational. All burning ammunition.

The two competing reads of where price should be

Anas Alhajji published a corrective on May 27 arguing that commercial crude inventories are not collapsing. He’s right on the data. The EIA report shows commercial crude (excluding SPR) at 445.0 million barrels — down 7.9 million week-over-week but up 1.9 million year-over-year. The “tank bottoms imminent” framing analysts have been deploying conflates SPR drawdowns with commercial draws. Commercial inventories are not crashing. They are bleeding moderately while SPR carries the load.

This matters for the organic shadow price calculation. A prior deep dive estimated organic WTI at $150-155 against the $90 paper print, implying a $60 suppression delta. Alhajji’s correction — commercial inventories holding up better than the consensus has been pricing — moves the realistic organic shadow price closer to $115-130. Still a substantial gap against the $90 paper print. Still meaningful for the eventual snap. But more moderate than the $150+ framing assumed.

Three things survive Alhajji’s correction:

1. The SPR drawdown rate is real and accelerating. From 1.229 mb/d the prior week to 1.417 mb/d this week. Treasury is burning ammunition faster, not slower.

2. The aggregate suppression apparatus is exhausting on the same timeline. Floor breach at the 243-million-barrel statutory defense reserve happens in late August at current pace.

3. Treasury, State, and Energy are engineering this for a specific landing — the September-October 2026 chip-sector cascade triggered by the helium scarcity hitting Asian fab earnings disclosures.

The snap is coming. It’s $115-130, not $150-160. Moderate is the new violent.

The Natty Greenback hiding in the data

The most important number in this week’s EIA report isn’t the SPR draw. It’s the U.S. crude export figure: 5,604 mb/d, up 59 percent year-over-year. The U.S. is approaching net-export status on crude. Cheniere, Venture Global, Sempra, and the Permian operators are already cashing the checks while the front-month WTI print appears suppressed.

This is the Natty Greenback in real-time data. The U.S. domestic price is being held low partly through aggressive export channeling — moving the crude to Europe and Asia at JKM-adjacent prices while U.S. refiners get the suppressed Henry Hub-adjacent input cost. European utilities and Asian refiners pay the spread. American producers and LNG operators collect it.

For all the focus on the paper-market suppression mechanics, the physical reality is that the U.S. is exporting its way through this with allied economies paying for the privilege. Someone is bearing the cost of the suppression. It’s not Washington and it’s not the Permian basin.

What China can do, and why every option costs something

China is in a pincer. The U.S. apparatus can push prices down (current posture) or release the suppression on its preferred timing. Either move costs Beijing.

Path A: stay passive on demand. China gets short-term cheap crude inputs. Domestic petrochemical complex margins stay healthy. Strategic reserves continue filling at low prices — Beijing has added roughly 169 million barrels of SPR capacity through 2025-2026, and the buildout continues. This is the current Chinese posture.

But Washington’s preferred timing for the eventual snap is September-October 2026, when the chip-sector cascade lands and the broader rotation absorbs the inflation shock. That cascade hits Chinese semis — SMIC, Huawei, the entire indigenous chip stack — and the global goods-demand environment Chinese exports depend on. The cheap oil today is being paid for by structural damage to the Chinese tech sector and export engine six months from now.

Path B: accelerate consumption to force the snap. Buy aggressively now to drain the inventory buffer, push prices toward organic shadow, force the rubber band to snap in July or August before the chip cascade lands. The U.S. faces a stagflation scenario the Fed cannot manage cleanly. The rotation gets disrupted. Washington’s timing breaks.

The cost to Beijing is steep. The marginal additional demand needed to overwhelm the U.S. apparatus is large — SPR alone has 130-150 million barrels of usable drawdown left at current pace. Chinese demand surge has to be massive, and gets paid for at increasingly high prices. China would also face U.S. response options: escalated Iran secondary sanctions cutting Chinese teapot refiners off from cheap Iranian crude, accelerated chip controls, coordinated Saudi or UAE surprise output boost to defuse the spike. The pincer can tighten.

And stagflation hurts China harder than the U.S. China is more physically exposed — refining, petrochemicals, transport, fertilizer. The U.S. is more financially exposed — Fed credibility, equity multiples. The U.S. has a rotational receiving basin (Genesis Coalition firms: Cheniere, advanced nuclear, MP Materials, the industrial reshoring base) to catch the rotating capital. China has fewer obvious rotational receivers.

The soft-power play and why Washington wants it broken

Beijing’s demand restraint isn’t just hedging. It’s a regional soft-power move. China is the indirect counterweight to high global oil prices for Japan, South Korea, India, and the ASEAN bloc — Asian importers whose industrial bases are far more vulnerable to crude spikes than China’s coal-anchored economy. Every barrel China doesn’t buy at the margin is a barrel that stays available to those importers at suppressed prices. Beijing gets to position itself as the reasonable regional actor maintaining stability, while the U.S. burns down its strategic reserves and runs head-fake diplomacy through Treasury and State.

This is the soft-power story the OilPrice.com piece gestures at. It’s real, and it matters more than the article credits.

It also matters more than U.S. strategy can tolerate.

The Anaconda strategy was designed as Chinese containment — two decades of U.S. naval posture in the Persian Gulf and Malacca, calibrated to choke off seaborne energy imports to Beijing. China built around it: coal-anchored grid running on domestic supply, overland pipelines from Russia and Central Asia, strategic-reserve buildout. The doctrine no longer functions as direct containment of China. What it functions as now is a leash on U.S. allies. Chokepoint pressure on Middle Eastern oil flows hits Tokyo, Seoul, Taipei, and the broader Asian importing bloc far harder than it hits Beijing. Washington’s preferred outcome for those allies is structural energy stress that locks them tighter into U.S.-priced LNG contracts, Henry Hub indexation, and the longer dependency arc that channels their industrial value into the U.S. export complex.

Chinese restraint disrupts that. By holding marginal demand off the market, Beijing keeps crude prices low and gives Asian allies room to breathe without leaning harder on U.S. supply. The soft-power dividend is real.

The U.S. response is structural. The current suppression apparatus — SPR drawdowns, talk-down statements, the dealer-gamma ceiling — keeps the suppressed paper price low enough that China gets no credit for its restraint. Asian allies don’t perceive Beijing’s hand because the paper market doesn’t reflect a high price for Beijing to have held down. The soft-power signal gets buried in the noise of a suppressed market.

When the apparatus exhausts and the snap arrives, Asian allies face a sudden re-pricing event Beijing cannot shield against. China’s strategic reserves can buffer China. They cannot buffer Japan or South Korea. The post-snap energy environment is exactly the structural stress the Anaconda doctrine was designed to produce — and it produces it on Washington’s timing, against allies who will have nowhere to turn except long-term contracts with Cheniere, Venture Global, and Sempra on the terms those operators are willing to offer.

Washington isn’t merely tolerating Chinese restraint. It is actively neutralizing the soft-power dividend Beijing is trying to bank, and engineering the snap to land in a way that hands Asian allies to the U.S. supply complex on terms they cannot refuse.

The smart Chinese move is still a losing move

Beijing’s observed behavior is the sophisticated play: stay passive on import surge, build strategic reserves aggressively, keep the petrochemical complex margins healthy on cheap inputs, and position the SPR to absorb the eventual snap on China’s terms rather than panic-buying into a violent re-pricing. The teapot refiners pulled back from Iranian crude after the May 1 OFAC sweep but the broader strategic accumulation hasn’t slowed.

This is not cooperation with U.S. strategy. It is hedging — the least-bad move available given the pincer. But it doesn’t escape the pincer. It just buys insulation against the snap when it comes.

The suppression apparatus is calibrated to pin Beijing into the option that is worst for China long-term: cheap inputs now, structural damage to the tech sector and export engine in the fall. China’s hedge softens the inflation shock when it lands. It does nothing to defuse the chip-sector cascade or the goods-demand collapse that the broader rotation requires.

The Kissinger payoff

Washington’s posture toward each category of state is now visible in the price action.

To be an enemy of America is dangerous. Iran is being administratively starved through the EO 13902 OFAC sweep, the rial collapsed to 1.8 million to the dollar, Iranian oil revenue is down $170 million per day.

To be an ally is fatal. European refiners are paying JKM-adjacent prices for crude they previously sourced cheaply via Hormuz, funding the U.S. export complex that is now hollowing out their industrial base. South Korean petrochemical operations face feedstock shortages that flow directly into export-margin compression. Japan, Taiwan, Germany — all paying. Chinese restraint provided some quiet relief while the suppression held. The post-snap environment removes even that buffer.

To be the smartest non-aligned power is still expensive. Beijing is hedging optimally and still getting pinched into financing its own future inflation while watching its tech sector get crashed on Washington’s preferred timing.

The pincer doesn’t have an exit. It has a least-bad option. China is taking it. Washington is fine with that.

The timing instrument

Next Wednesday at 10:30 AM Eastern, the EIA releases the next Weekly Petroleum Status Report. Two numbers will tell you where this is going.

First: does the SPR drawdown rate hold above 1.4 mb/d, or does Hegseth at Defense start decelerating to defend the 243-million-barrel statutory floor? Acceleration means the apparatus is committed to the current timing. Deceleration means Treasury is starting to ration ammunition for the late-August window.

Second: does the crude export figure hold above 5.5 mb/d? If yes, the Natty Greenback channel is doing its work and U.S. domestic suppression is being underwritten by allied energy payments. If exports drop, the export-channeling component of the suppression is weakening.

Watch those two numbers. They will tell you whether Washington is still on the preferred timing — chip cascade lands first, oil snap absorbs the inflation, rotation completes — or whether the timing is starting to slip.

The eighth Iran-deal head-fake today dropped WTI another four dollars on identical mechanics to the prior seven. The diminishing-returns pattern is visible: each successive head-fake produces a smaller amplitude and faster mean reversion. The apparatus is degrading. The question is whether it lasts until September.

The pincer is U.S. strategy asking Beijing to choose how it bleeds. Beijing is hedging. Washington is fine with that. The deadline is real.

Here’s the arithmetic. Technology stocks now constitute approximately 30% of the S&P 500 by index weight. The energy sector sits at roughly 5-6%. The Genesis Architecture identifies energy, defense, semiconductors, biomanufacturing, and frontier industrial capacity as the sectors that have to be capitalized at scale.

If the index reweights from a tech-heavy composition to an energy-and-industrial composition, the math gives exactly two options. Option one: tech holds at current valuations while energy multiples expand by a factor of five or six. That has no historical precedent. Energy stocks at 50-70 P/E ratios while the underlying commodities trade at modest multiples is fantasy. Option two: tech multiples compress hard while passive money mechanically flows into the sectors that benefit from the compression. That’s the BlitzCorrection thesis I have been writing about since Feburary 2026 (but thinking about, and kicking tires on in lectures for much longer).

There is no third option at current sectoral weighting. The plumbing on this won’t allow it to be slow (more on this below)

The high-margin problem (again)

The reason the rotation can’t be slow is that the rentier coalition’s margin structure is too extreme to compress gradually. NVIDIA gross margins at 74% aren’t normal corporate profitability. That’s wannabe cocaine kingpin economics with a NASDAQ ticker. Apple’s App Store toll, Meta and Google’s ad monopolies, the pharma-PBM oligopoly’s spread between manufactured cost and reimbursed price, the legacy defense primes’ cost-plus contracts that reward inefficiency and punish speed: rents masquerading as margins. They don’t compress on a sliding scale. They get broken.

The rentier coalition built itself on a specific arrangement: a Federal Reserve that suppressed correction mechanisms for fifteen years, a passive-bid asset complex that mechanically purchased index components by market cap, and a regulatory environment that captured its own enforcers. That arrangement produced the highest concentrated profit margins ever recorded. You don’t unwind that gradually. You unwind it the way you unwind a cartel: by removing the protection and letting competitive pressure crush the spread.

The Genesis Mission requires cheap inputs. Cheap chips for autonomous defense systems. Cheap energy for AI compute. Cheap drug discovery for biological sovereignty. Cheap manufacturing for industrial reshoring. None of those inputs can be cheap while the rentier margins stay where they are. The rotation is a margin demolition. It’s why I am thinking about deflation so much these days.

How crashes used to work

Again, i’m on repeat: Throughout American financial history, the standard pattern was simple. Markets concentrated into a dominant sector. The dominant sector compressed. Capital fled to cash and gold and short-duration Treasuries. The economy contracted. New sectors slowly emerged through entrepreneurial discovery and capital reallocation over years or decades. 1929 produced the depression and a fifteen-year wait for the postwar industrial boom. The 2000 dot-com crash bled into 2003 and the dominance pattern that emerged after didn’t fully establish itself until the early 2010s. 2008 produced six years of zero-rate suppression before any genuine sectoral repricing got allowed.

The slow-emergence pattern was the only available path because there was no pre-coordinated successor architecture. The market had to discover what came next after the crash. That discovery is expensive in time, capital, and political stability.

That pattern doesn’t apply anymore.

What’s different now

Four things have changed since the last time American capital had to go through a major sectoral rotation, and they’ve changed in ways that compound.

The first change is algorithmic. The plumbing of the equity market in 2026 will produce compression at speeds that didn’t exist in any prior cycle. Four pieces of that plumbing matter.

The Market-on-Close auction (MOC) is the daily 4pm window when the exchange matches all “trade at the close” orders at a single official closing price. It now handles roughly 10-20% of total daily NYSE volume, concentrating a huge fraction of the day’s price discovery in a thirty-second mechanical window dominated by index funds rebalancing to track their benchmarks. When wrong-way flows hit that close, there is no discretionary capital left in the room to absorb them.

Zero-day-to-expiration options (0DTE) are option contracts that are bought and expire on the same trading day. They now make up roughly half of all S&P 500 options volume. The bank dealers who sell those contracts have to hedge their positions by buying or selling the underlying stock, and as expiration approaches the hedging flows accelerate in whichever direction the market is already moving. What used to be a multi-day options-expiration dynamic now collapses into hours, amplifying whatever the intraday move already is.

The historical “glass floor” was the price at which discretionary buyers — pension funds, value managers, retail investors — would step in to support a falling stock. The floor would lower with the price, but it stayed in place. The modern equivalent doesn’t lower. It vanishes. Passive funds and high-frequency market makers bid mechanically inside narrow stress tolerances, and once stress crosses those thresholds the bid disappears entirely. Prices drop without anywhere to land.

High-frequency trading firms (HFT) — Citadel Securities, Virtu, Jane Street, and a handful of others — supply most of the moment-to-moment buy and sell quotes that make the market liquid during a normal trading day. Their algorithms cancel those quotes within milliseconds when their models detect stress. The liquidity the market most depends on at moments of crisis is precisely the liquidity that disappears at those moments, because the firms providing it are optimizing for not getting run over by the next move.

Stack those four mechanisms together and the velocity of both the crash and the recovery is structurally faster than 1929 or 2000 or 2008 by orders of magnitude. Compression that took months historically can take days. Recovery that took years can take months.

The second change is policy. The successor architecture isn’t being discovered post-crash. It’s being legislated and deployed now, while the rotation is still in setup. The CHIPS and Science Act of August 2022 and the Inflation Reduction Act began this layer under the Biden administration, with $52.7 billion in semiconductor manufacturing subsidies and trillions in clean-energy industrial credits. The current administration has stacked five more sovereign instruments on top of that foundation.

The GENIUS Act of July 2025 regulates dollar-backed stablecoins (Tether, Circle, PayPal’s PYUSD) by requiring issuers to back every digital dollar with short-duration U.S. Treasury securities and prohibiting them from passing the yield on those Treasuries to customers. The effect is to convert the entire stablecoin reserve base into mandatory Treasury demand. Treasury Secretary Bessent has stated that the GENIUS Act will produce a structural surge in Treasury buying. From anecdotal Youtube/Social media finance viewing, I think many of the incumbent analysts are vastly underestimating how powerful this shift will be in maintaining Pax Dollar Hegemony.

The SPEED Act, included in the FY26 National Defense Authorization Act (the annual NDAA that sets Pentagon spending and procurement rules), expanded the dollar thresholds at which the Pentagon can buy commercial items off-the-shelf rather than custom defense systems built to bespoke specifications. The effect is to open the procurement pipeline to firms like Anduril and Palantir that previously couldn’t compete with Lockheed Martin’s contract-writing apparatus.

Executive Order 14363, signed November 24, 2025, formally launched the Genesis Mission: a federal program coordinating AI, biomanufacturing, frontier energy, and advanced manufacturing R&D into a unified industrial buildout. Energy Secretary Chris Wright has called the compute side of it the next Manhattan Project.

Section 232 Proclamation 11002 of January 14, 2026 invoked the national-security tariff authority of the Trade Expansion Act of 1962 to impose tariffs on imported advanced semiconductors. The effect is to compress margins on the existing chip layer (especially NVIDIA’s 74% gross margin) while protecting the domestic fabs being built with CHIPS Act money.

The BIOSECURE Act of December 18, 2025 restricts U.S. federal contracting with Chinese-linked biotech firms (notably BGI, WuXi AppTec, WuXi Biologics), reshoring biomanufacturing demand to the U.S. firms in the BioMADE / Ginkgo / Twist / Recursion cohort.

The SHIELD IDIQ of March 3, 2026 is the Pentagon’s $151 billion multi-vendor contracting vehicle for unmanned and AI-driven defense systems. (IDIQ stands for “indefinite-delivery, indefinite-quantity,” the procurement format that lets the Pentagon issue task orders to a roster of pre-approved vendors over a multi-year period without re-bidding each contract.) It is the funding pipeline for the Defense-Industrial 2.0 cohort.

All of these are sovereign instruments that pre-stage the receiving sectors.

The third change is capitalization. The receiving basins aren’t speculative. They’re pre-capitalized through corporate balance sheets, sovereign wealth deployment, and direct industrial CapEx that’s already in the ground.

TSMC Arizona ($165 billion), Samsung Texas ($230 billion), and Micron New York ($100 billion) are operating semiconductor fabrication plants (”fabs” in industry shorthand) with real revenue, manufacturing leading-edge chips on U.S. soil. Anduril’s Arsenal-1 facility in Pickaway County, Ohio breaks ground for production in July 2026. The Pentagon’s AI Acceleration Strategy, launched January 2026, routes procurement away from the legacy defense primes (Lockheed Martin, Northrop Grumman, RTX, Boeing Defense, General Dynamics) and into the Defense-Industrial 2.0 vendor stack (Anduril, Palantir, SpaceX, Shield AI, Saronic, Epirus).

The Stargate consortium — OpenAI, Oracle, SoftBank, the UAE sovereign tech vehicle MGX, with CoreWeave and Crusoe at the infrastructure edge — has $500 billion committed and over $100 billion already deployed in AI compute infrastructure. U.S. hyperscaler capital expenditure for 2026 is guided at $720-750 billion, a 67% year-over-year increase, locked into multi-year industrial deployment of data-center and compute capacity. LNG export terminals on the U.S. Gulf Coast are running at 94% of nameplate capacity, with another 13.9 billion cubic feet per day (Bcf/d) of export capacity under construction.

Gulf state sovereign wealth funds were allowed to buy in early. The Microsoft-G42 IGAA Compliance Regime, signed April 2024, was the first bilateral framework integrating UAE AI infrastructure into the U.S. compute monopoly. (G42 is the UAE’s leading AI firm; IGAA — the Investment in Gulf and American Architecture — sets the cybersecurity and technology-transfer terms required for U.S. tech firms to operate at scale in the Gulf.) Saudi Arabia’s Public Investment Fund (PIF), the kingdom’s sovereign wealth fund, committed $100 billion to HUMAIN, the AI subsidiary PIF launched in 2025. Qatar’s Qatar Investment Authority (QIA), the Qatari sovereign wealth fund, struck a $20 billion joint venture with Brookfield, the Canadian asset manager, to fund U.S. industrial assets. MGX took its position inside the Stargate consortium itself. Treasury and the National Security Council coordinated those allocations to lock the Gulf into the new architecture before the rotation became visible. Those funds can’t easily exit even when Iran war tensions strain the relationship: the dollar system and SWIFT (the global interbank settlement network) are the only viable rails for surplus oil revenue, and gold absorbs only so much.

The fourth change is the passive bid. Every prior sectoral rotation required active capital — fund managers, individual investors, pension committees — to make discretionary choices about where to redeploy. The 2026 rotation has roughly $30 trillion in passive index assets that mechanically reallocate into whatever the index composition turns into. BlackRock, Vanguard, and State Street don’t have to decide that Defense-Industrial 2.0 deserves more weight than Lockheed Martin. The index decides. The funds follow.

Underneath that sits a continuous mechanical inflow from 401(k) auto-deductions and target-date funds, which together hold $5.2 trillion in assets under management (AUM) and are growing at 21% year over year. Target-date funds are retirement portfolios that automatically rebalance toward bonds as the saver approaches retirement age. They are the default option in most U.S. employer-sponsored retirement plans, which means most U.S. workers have their retirement savings on autopilot inside them. That flow buys whatever the index contains, on every payroll cycle, regardless of what the market is doing.

Stack those four together and the slow-emergence pattern is geometrically impossible. The crash compresses faster than any prior cycle. The policy backstop is already legislated. The receiving sectors are already capitalized. The recovery bid is already mechanical. What used to take a decade can collapse into months at the index level.

The lifeboat was already built (i.e. new vessels to receive capital)

This is the biggest historical structural difference that matters most. In 1929, when RCA and GM compressed, there was no place for capital to flow except cash and gold. The receiving basins didn’t exist. Capital sat in short Treasuries for years while Roosevelt’s administration improvised the New Deal. In 2000, dot-com compression had no productive successor either; the receiving basin turned out to be the housing bubble, which compressed eight years later. In 2008, the receiving basin was the Federal Reserve’s own balance sheet through quantitative easing. The Fed sat on the bid for fifteen years and capital allocation stalled.

The 2026 BlitzCorrection has receiving basins that are already operating businesses, with revenue on the books and capital expenditure in the ground before the crash starts.

When tech multiples compress, capital doesn’t have to discover where to flow. The infrastructure is already standing up. The receipts are documented. The contracts are signed. The sovereign capital is committed. The industrial base is breaking ground.

Every other major rotation in American financial history had a vacuum on the other side. This one has a fully capitalized industrial program waiting to absorb the rotation.

What follows from this

The math forbids slow. Would American policy planners do a slow rotation if they could? Possibly. I don’t think they are going to force a blitzcorrection because they WANT to, but will do it because the political imperative leaves them no other choice. The Iran war serves this function. It allows American power to keep the strait closed, throw sand into the gears of the global economy—while the propaganda machine will all but inevitably write this off as a ‘blunder’ ‘mistake’, ‘tactical error’ or whatever. Again, I don’t want to sell this as 3d chess. But for anyone who thinks the U.S. would never undermine their own economic order to gain foreign policy or domestic advantage—see the 1979 shift to Paul Volcker at the Fed, and the rate hiking inflation busting bonanza that followed, or the 1973 Yom Kipper war that was pursued along similar lines—with the latter case providing the best parallel to what’s happening now, namely the power politics move combined with internal propaganda portraying the policy as a disaster that lead to gas rationing and stagflation.

Ultimately, the historical pattern of crash-then-slow-discovery doesn’t apply because the conditions that produced that pattern have been replaced by their opposites. Compression will be fast. Recovery will be similar velocity. The composition of what gets rebuilt won’t match the composition of what gets compressed.

Whether all of this plays out as a clean V-shaped index recovery or something messier depends on variables that haven’t fully resolved yet: Genesis CapEx returns showing up by 2027-2028, the Federal Reserve’s willingness to let velocity run rather than suppress it, sovereign capital staying committed across the transition, the absence of an exogenous shock that breaks the algorithmic plumbing in unmanaged ways. The base case is a discrete repricing event of unusual velocity, followed by a recovery of similar velocity, at a sectoral composition that bears little resemblance to the one that compressed.

That outcome is structurally distinct from anything in the historical record. The crash that’s coming can’t be like the last ones because the architecture surrounding it isn’t like the last ones. The plumbing has changed. The policy has changed. The receiving sectors have changed. The bid mechanism has changed. The next piece in this series will turn to what that means for who actually pays for the rotation.

Financial factions that dominate industrial factions are not unique to the United States, a brief detour into the British POV of the rise of the U.S. as an industrial powerhouse—purportedly through the use of “ruinous competition” that sounds alot like “overcapacity”.

In October 1903, Joseph Chamberlain stood up in Liverpool and gave a speech that has been giving its lines to U.S. Treasury Secretaries ever since. Chamberlain was the most powerful figure in the British Conservative Party’s industrial wing, the architect of the Tariff Reform campaign, and at that moment the loudest voice in British politics demanding a defensive economic response to a rising industrial power eating Britain’s lunch. The rising power was the United States. The lunch was British exports of steel, machinery, agricultural goods, manufactured products of every kind. Chamberlain’s speech that day defined the pejorative term that would dominate the discourse of declining hegemons for the next 120 years.

“Dumping is the placing of the surplus of a home manufacture in a foreign country without reference to its original and normal cost. Dumping takes place when the country which adopts it has a production which is larger than its own demand. Not being able to dispose of its surplus at home, it dumps it somewhere else.”

That was Chamberlain, in 1903, defining “dumping” for the British public. The country he was talking about was the United States. The product he was talking about was American steel, American machinery, American manufactured goods piling up in British ports at prices British producers could not match.

The structural read of late-Victorian Britain was that this was unsustainable. American manufacturers, the British financial establishment said, were operating below cost, fueled by debt, propped up by ruinous competition that would inevitably collapse the entire American industrial economy under its own contradictions. The structural read was wrong. It was wrong every year for forty years. The British political establishment never recognized it was wrong until the First World War, when the sheer volumetric superiority of American industrial output became the determining factor in global survival, and by then it was too late.

The Americans were running what economic historians later named productive deflation. The U.S. was in the middle of a sustained deflationary period (1873–1896). The deflation was not destructive. It was driving capital accumulation. American industrial firms in the high-fixed-cost sectors (steel, rail, oil refining) kept their plants running at full capacity even when prices fell, because shutting down meant defaulting on debt payments and losing the plant. Running at full capacity drove unit costs down. Unit costs falling drove prices down further. American producers captured market share by being structurally cheaper at scale than anyone else. Bankruptcies happened. Equity holders got wiped out. The plants stayed open. New owners bought the physical capital for pennies on the dollar and kept producing. By 1900, the U.S. had passed Britain as the largest industrial economy in the world. By 1913, U.S. manufacturing productivity was 2.13 times the British level.

The British “ruinous competition” diagnosis was loud, sustained, and constant throughout this entire period.

Begin with the Royal Commission on the Depression of Trade and Industry, convened by Westminster in 1886 to address the deflationary slump in British manufacturing. The Commission collected testimony from across British industry on the causes of British stagnation. Witness after witness pointed to the United States. The British Embassy in Washington filed reports for the Commission cataloguing what it saw as American economic fragility:

“Signs of depression in the United States: a severe decline in the construction of railroads, factories and other works; reductions in the manufacture of cottons... iron and steel... and the consequent, lowering effects on employment and wage rates.”

The 1886 Royal Commission dwelled at length on these “signs of depression.” It interpreted American railroad bankruptcies, the frequent corporate failures of the deflationary panics of 1873 and 1893, and the rate wars between American railroads as evidence that the American competitive surge was a transient phenomenon that would shortly burn itself out. The Commission failed to recommend the systemic industrial restructuring required to match American scale. To do so would have required acknowledging that British Free Trade orthodoxy was failing British industry. The British rentier class, meaning specifically the City of London financial bloc that benefited from cheap foreign imports and global capital mobility, was politically dominant and would not accept that conclusion.

The British financial press carried the diagnosis through the 1890s and 1900s. The Economist, The Times, the Commercial and Financial Chronicle all dwelled on the chaotic nature of American capitalism. City of London publications used “ruinous competition” as a stock descriptor for American practices. American steel exports flooding European markets at prices below British cost of production were framed not as evidence of American productive superiority but as unnatural anomalies born of protectionist manipulation and desperate American overproduction. In 1900, the Commercial and Financial Chronicle eagerly noted when “the ruinous competition in prices which formerly existed had been ended,” reflecting British financial anxiety and the City of London’s deep desire for American consolidation that would stabilize the price level and protect rentier-friendly margins.

The specialized British trade journals carried similar judgments from inside British industry itself. The Iron and Steel Trades Journal, Engineering, and other industrial periodicals treated American manufacturing methods with a mix of grudging admiration and strategic dismissiveness. American factories, they wrote, ran machinery at maximum speed until it was destroyed, then scrapped it. American practices were wasteful, unscientific, and financially unsound compared to British craft traditions. The British engineering establishment, bound by a culture of durability, steady returns, and gentlemanly capital allocation, did not understand that the rapid depreciation of American capital equipment was a deliberate feature of an economy optimizing for continuous technological substitution and maximum throughput. The Americans were running productive deflation at industrial scale. British engineers thought it was carelessness.

By the turn of the twentieth century, the British public was reading what came to be called the “American Invasion” literature. Three books defined the genre, all published in 1902. F.A. McKenzie’s The American Invaders documented how American manufactured goods were aggressively displacing British products from Madrid to Saint Petersburg. McKenzie wrote with a particular bitterness about what he saw as the imperial irony of the situation:

“British blood bought these new dominions... paid the price to benefit the traders of other lands, and... Americans especially will reap the commercial profits of our triumph.”

W.T. Stead’s The Americanization of the World; or, The Trend of the Twentieth Century carried a more apocalyptic note. Stead recognized the structural shift but read it as inevitable:

“We shall descend slowly but irresistibly to the position of a second-rate Power... The Briton, instead of chafing against this inevitable supersession, should cheerfully acquiesce in the decree of Destiny, and stand in betimes with the conquering American.”

B.H. Thwaite’s The American Invasion, written by a British engineer with technical training, scrutinized American methods more analytically. Thwaite identified the reliance on massive capitalization, continuous technological substitution, and economies of scale. He concluded the American factory floor was a “menace,” a fragile system overheating and destined to collapse under the weight of its own excess production. It would not collapse. The American factory floor would sustain itself for the next century. The British factory floor would not.

Chamberlain’s 1903 Tariff Reform campaign was the political culmination of the British misreading. Chamberlain, having watched his country’s industrial position deteriorate for thirty years, broke with British Free Trade orthodoxy and demanded a defensive imperial preference system, meaning a tariff wall around the British Empire to insulate British industry from American and German competition. His Liverpool speech defined dumping. His subsequent speeches across the country attacked the U.S. as an unfair competitor:

“We are the only country that keeps open ports. All the other great countries protect themselves against dumping... Nothing will prevent the people of this country from immediately imposing a duty which shall defend against such unfair competition a great and staple industry.”

The Free Trade versus Tariff Reform debate of the 1890s and 1900s was, in retrospect, the British political establishment’s only serious chance to confront the structural

shift. Chamberlain’s bloc argued that American productive deflation was destroying British industry and the British working class. The Free Traders, representing the financial capital faction and the City of London, argued that cheap American goods benefited the British consumer, and besides, the American trusts were monstrous monopolies that would inevitably collapse under their own internal contradictions.

Chamberlain’s movement failed. The Conservative Party fractured. The Liberals won the 1906 general election. British Free Trade orthodoxy held. The City of London continued to direct British capital to higher-yielding investments abroad rather than to British industrial reinvestment. British industry continued to atrophy. By 1914, when the British and German empires went to war and discovered that the entire war would be decided by industrial output capacity, Britain was already structurally dependent on American manufacturing to fight it. The empire’s transition from sovereign hegemon to subordinate partner, eventually formally a U.S. client state, was already running. The British political establishment recognized the depth of its forty-year misreading only retrospectively, primarily during and after the First World War, when the verdict had already been entered.

A reader scanning U.S. discourse on China in 2024 through 2026 will find the same vocabulary repeating. Janet Yellen at Treasury defining Chinese strategy as “dumping” and “overcapacity,” lifted whole from Chamberlain’s 1903 Liverpool playbook. The USTR Section 301 reports declaring Chinese practices “environmentally unsustainable and economically unviable for U.S. producers,” which is the exact structural diagnosis the 1886 British Royal Commission delivered on American producers. The Heritage Foundation reports on how China is “hamstrung by an increasingly poor regulatory and business environment and unsustainable levels of spending and debt.” The Wall Street Journal and Financial Times consensus framing Chinese capacity expansion in solar, EVs, batteries, and semiconductors as a desperate attempt to export its way out of a domestic recession. The script was written in London 122 years ago. The U.S. is reading from it line by line.

The structural read of declining hegemons is that they cannot see what is happening to them, because seeing it would require admitting that the financial-rentier configuration of their own economy is the disqualifier. The British rentier coalition could not admit, in 1903, that the City of London’s free-flowing capital had bled British industry of the reinvestment that would have allowed it to compete. The U.S. rentier coalition cannot admit, in 2026, that thirty years of antitrust non-enforcement and financialization gutted the American industrial base it now wants to rebuild on the cheap.

The vocabulary repeats because the structural position repeats. The British oligarchical infighting was won by the finance and free trade wing. The U.S. was reading from the same script, with the same vocabulary, with the same blind confidence that the rising power’s collapse is just a quarter or two away, for the past 30 years.

Fat chance.

In February 2026, the University of Michigan Consumer Sentiment Index sat at 56.6. That is near historic lows. The reading was 13 percent below the year-ago level and roughly 21 percent below January 2025. Forty-six percent of survey respondents spontaneously reported that high prices were eroding their personal finances. Inside the same month, the S&P 500 hit successive all-time highs. The Buffett Indicator (total U.S. market capitalization to GDP) reached 229.5 percent, roughly 2.4 standard deviations above its long-term historical trendline. The Shiller CAPE ratio sat near 41.83, an extreme historical premium. Real retail sales had been below their long-term trendline continuously since July 2022.

These data points come from the same country. They do not come from the same economy.

The University of Michigan release explained the divergence inside its own data. Sentiment among large stockholders rose during the period. Sentiment among consumers without stock holdings fell. The aggregate index masks a class-divided reading. Asset-owning America was feeling fine. Wage-earning America was feeling like it was being financially destroyed. The S&P 500 prices the first one. The Consumer Sentiment Index measures the second one.

The structural read is that these two Americas no longer share a single economy. They share a country, a tax system, a flag, and a currency. They do not share an economic experience. The asset-owning country has detached from the wage-earning country, and the U.S. equity market is the layer at which the detachment is most visible.

The disconnection has a third dimension that the first two layers obscure. Even if the U.S. equity market did its theoretical job (price discovery for productive enterprises) and even if asset-price growth tracked real-economy growth, the gains would still accrue to a small fraction of Americans. The asset class itself is structurally concentrated.

Federal Reserve data from late 2025 shows the top 1 percent of U.S. households owns 31.7 percent of total household wealth, holding approximately $55 trillion in assets. That figure roughly equals the combined wealth of the bottom 90 percent of the country. The top 10 percent owns more than 87 percent of all corporate equities and mutual fund shares. Roughly 87 percent of Americans who own stocks at all live in households earning over $100,000 a year. The bottom half of the country owns somewhere between 1 and 3 percent of total household wealth and a vanishing share of equities.

About 58 percent of Americans report owning stocks at all, mostly through employer-sponsored retirement accounts. The median 401(k) balance among those who have one sits in the $35,000 to $40,000 range. That is not exposure to the asset class in any structurally meaningful sense. It is enough to feel the volatility on the way down. It is not enough to participate in the way up.

So when the financial press reports that “the market hit a new high,” the implicit subject is the asset-owning decile. For the rest of the country, the market hitting a new high is something happening to other people’s portfolios. The S&P 500 is not the American economy at the level of corporate finance. It is not the American economy at the level of price discovery. It is also not the American economy at the level of who actually owns it.

Three layers of disconnection. The architecture is not subtle.

The numbers are not subtle either. From 2000 to 2026, the cumulative return on the S&P 500 ran roughly 632 percent. Cumulative nominal U.S. GDP growth over the same window ran approximately 200 percent. Asset prices grew at more than three times the rate of the underlying economy. The relationship between asset values and productive output is not “imperfect” or “noisy.” It is structurally severed. The S&P 500 is not pricing the U.S. economy. It is pricing something else. And that something else is owned by ten percent of the population.

What it is pricing becomes clearest at the sectoral level.

Take the global electric vehicle market. In 2023, BYD delivered 2.26 million EVs globally, formally surpassing Tesla, which delivered 1.64 million. By 2026, Chinese manufacturers dominated the top 20 global EV sales charts. BYD’s vertically integrated supply chain and aggressive cost compression had turned it into the largest EV producer in the world. By any measure of the actual auto industry, BYD was the dominant firm.

The U.S. equity market did not notice.

In early 2026, Tesla traded at a forward P/E ratio above 400. BYD traded at a forward P/E around 18 to 23. Tesla’s market capitalization was roughly $635 billion. BYD’s was roughly $89 billion. Tesla was worth about seven times more than the company that was outproducing it by half a million vehicles a year and growing market share faster across more geographies.

The U.S. market was not pricing the cars. The U.S. market was pricing a sentiment narrative about robotaxis, autonomous AI, and Elon Musk’s branding. BYD was producing the actual industrial reality of the global EV transition. Tesla was producing a story. The story was worth seven times the reality. That is not a price-discovery mechanism. That is a sentiment-storage system with a stock ticker on top.

The same disconnection runs through the rest of the U.S. tech complex. The hyperscalers (Amazon, Alphabet, Meta, Microsoft) committed roughly $650 to $700 billion in AI capital expenditures across 2024 to 2026, which the equity market priced at perfection while ignoring the question of whether AI end-user revenue would ever justify the spend. Palantir traded at 227 times trailing earnings and 71 times revenue while the CEO expensed a $17.2 million private aircraft and the company posted $684 million in dilutive stock-based compensation. The stock kept going up. The algorithms read the press releases the firm was configured to feed them. The bid did the rest.

Three layers produce this result.

The first is the passive bid. In 2025, exchange-traded funds drew in $1.5 trillion in inflows, the largest year on record, with passive indexed assets formally surpassing actively managed assets for the first time in history. BlackRock, Vanguard, and State Street index the largest companies regardless of whether the largest companies deserve to be the largest. Auto-deducted retirement contributions flow into the bid every two weeks, mechanically, price-insensitively. Long-term index funds took $74 billion in net inflows in March 2026 alone. Active funds bled $26 billion in net outflows over the same period. The mechanical bid dominates the price-setting layer.

The second is the buyback layer. Since the SEC’s Rule 10b-18 liberalized open-market share repurchases in 1982, U.S. nonfinancial corporations have been net buyers of their own equity, not net issuers. JW Mason at the Roosevelt Institute documented the inflection: in the 1960s and 1970s, an additional dollar of corporate borrowing yielded roughly forty cents of fixed productive investment. After the shareholder-revolution of the 1980s, that figure collapsed to less than ten cents. The remaining ninety cents went to dividends and buybacks. William Lazonick called the shift the move from “retain and reinvest” to “downsize and distribute.” In 2021 alone, S&P 500 firms spent $882 billion on buybacks. The empirical literature suggests roughly 75 percent of that volume was executed purely to inflate share prices rather than for any productive purpose. The 2020s have been running closer to $1 trillion annually. The corporate sector is not financing operations through equity. It is extracting capital from itself and transferring it to the asset-owning class through buyback-driven EPS engineering.

The third is the algorithmic sentiment overlay. Automated systems dominate trading volume in liquid markets. They parse the inputs they are configured to read: Reuters, Bloomberg, the wire services, U.S. Treasury statements, Truth Social posts from accounts the algorithms recognize as market-moving. They do not parse foreign-language industrial reports. They do not read Persian-language Telegram channels. They do not synthesize qualitative geopolitical reality. When Trump posts a Truth Social claim that Iran is near a peace deal, the algorithms move the market. When Iran simultaneously publishes Hormuz transit terms specifying reparations from the U.S., the algorithms do not notice. The pricing layer responds to inputs the system was built to read, and ignores inputs it was not.

Together, these three layers produce a market that mechanically ratchets asset prices upward, extracts corporate cash flow into asset-owner pockets, and responds to whatever sentiment input flows into the configured pipeline. This is what the U.S. equity market does. It is not what the U.S. equity market is supposed to do, but it is what it is structurally engineered to do, and it has been doing it for forty-some years.

China is on the other end of this comparison. Michael Pettis at Peking University frames the difference cleanly: the U.S. views investment as capital seeking returns, with the goal of maximizing short-term financial productivity. China views investment as a tool for capturing global market share and building national capacity. China force-feeds investment into the real economy, accepting lower rates of return to dominate physical production. The Chinese banking system, anchored by the Big Four state-owned banks (ICBC, China Construction Bank, Agricultural Bank of China, Bank of China), allocates capital by industrial-policy criteria rather than by quarterly shareholder yield. The Chinese equity markets exist, but they are subordinate to industrial policy. Buybacks are not a structural feature. Capital is not permitted to extract from the productive base for the benefit of the asset-owning class. The state does not allow it. This is one of my IR comparative maxims about U.S. and China politics: In the U.S., factions of capital competes to control the state, in China, factions of the state compete to control capital.

The result is BYD outproducing Tesla. The result is Chinese inference at fifty cents per million tokens against U.S. hyperscaler API at fifteen dollars (because…gotta get that IPO money). The result is a Chinese economy structurally organized to win on production while the U.S. economy is structurally organized to compress capital into asset prices that ten percent of Americans own.

The S&P 500 at all-time highs does not tell you that the American economy is doing well. It tells you that the wealth-storage layer of the American economy is doing well, because it has been structurally insulated from the productive economy underneath it. The two are connected only at the political level. The political coalition that owns the assets controls the policies that maintain the architecture, and the architecture is doing exactly what it was built to do.

Look at consumer sentiment. Look at real retail sales. Look at the BYD numbers. Look at what gets built, by whom, and where. Asset prices are the last thing to know. the people who own the market are not part of the solution—because they are part of the problem.