THE ECONOMY

The oil shock arrives in the real economy, and an airline is its first casualty

Spirit Airlines ceased operations this fortnight, unable to survive the fuel prices the war unleashed, as the oil shock moved from the markets into the machinery of ordinary life. JPMorgan warned that oil inventories could run dry within weeks. The war has stopped; its economic weather has only reached the ground.

The war's economic shock reached the ground this fortnight, and it took an airline first. Spirit Airlines ceased operations, its low-cost model unable to survive fuel prices that the war had doubled and a bailout that did not come, the first large corporate casualty of an oil shock that had, until now, lived mostly in the abstraction of the futures markets. It will not be the last. JPMorgan warned that developed-world oil inventories could deplete within weeks; Dubai's airport reported passenger traffic down two-thirds; gas prices climbed at pumps far from any front; and the shock that the ceasefire was supposed to relieve revealed itself to be embedded in the system, working its way, with the lag that always separates a financial shock from a human one, into jobs and prices and the ordinary calculations of ordinary households. The war stopped in April. Its economic consequences were, this fortnight, just arriving.

THE TRUCE

The ceasefire holds, barely, as the Strait stays contested

The Pakistan-mediated ceasefire held through the fortnight, fraying but intact, as Iran floated further proposals through intermediaries and the United States announced a naval operation to escort neutral shipping through the still-dangerous Strait of Hormuz. Sporadic strikes continued; each side accused the other of violations; and the Strait, through which a fifth of the world's oil must pass, remained a place approached with dread, its contested waters keeping the oil price elevated and the economic shock alive long after the largest guns had fallen quiet.

IN BRIEF

The models, the pumps, the bet

Recession models stayed near the threshold that has historically preceded a downturn, the oil shock and the tariffs together pushing the odds higher than they had been in years. Central banks confronted the supply-shock inflation their tools cannot address. And the great question hanging over the economy, whether the vast artificial-intelligence build-out could withstand a genuine macroeconomic shock, moved from the theoretical to the immediate, as the first recession in the AI era loomed and no one could yet say whether the bet was infrastructure or speculation.

The Column

Installation, and the crash that sorts it out

The AI build-out faces its first recession, and everyone wants to know if the bubble will burst. This is about a theory of how technological revolutions actually unfold, because it suggests the question is not whether there is a crash but what a crash is for.

By Daniel Reeve

The AI build-out is about to meet its first recession, and the debate has collapsed into a binary: bubble or not, burst or hold. I want to offer a more useful frame than the binary, drawn from the most illuminating theory we have of how technological revolutions actually move through an economy, because it suggests that the crash everyone is either dreading or predicting is neither the disaster the bulls fear nor the vindication the bears crave, but a specific and recurring phase with a specific function, and that understanding the function tells you more than any prediction of the timing.

Carlota Perez, studying the great technological revolutions of the last two centuries, from canals and railways to steel and electricity to the automobile and mass production, found in all of them the same structure. Each revolution unfolds in two great periods separated by a crash. First comes the installation period, when financial capital, exuberant and undisciplined, floods into the new technology, builds out its infrastructure at frantic speed and wild overvaluation, and produces a bubble, a frenzy of speculation that vastly overshoots the technology's near-term ability to pay. Then the bubble bursts, always, because the financial enthusiasm has run far ahead of the real economic value the technology can yet deliver, and the crash is savage and destroys fortunes. And then, crucially, after the crash, comes the deployment period, the long golden age in which the infrastructure that the bubble overbuilt, the railways, the fibre, the factories, having been laid down in the frenzy and then repriced to sanity in the crash, becomes the cheap foundation on which the technology's real and lasting value is finally built. The railways of the 1840s British mania ruined their speculators and then carried the goods of a century. The dark fibre of the dot-com bubble bankrupted its layers and then carried the internet that followed.

Now the honest counterargument, which the bulls will press and which has real force. This time, they will say, is genuinely different, because unlike the dot-com companies that had no revenue, the leaders of the AI build-out are among the most profitable enterprises in history, funding their construction from real cash flows rather than pure speculation, so the Perez pattern of a speculative installation bubble may simply not apply; there is no frenzy of valueless companies to purge, only profitable giants building real infrastructure. This is the strongest version of the case and it is not empty. The profitability is real and it is a genuine difference from prior manias. But notice what the frame reveals even so: the question is not whether the companies are profitable but whether the level of capital being deployed, the $690 billion a year, is justified by returns not yet demonstrated, and Perez's point is precisely that installation-phase capital always feels justified to those deploying it, always points to real technology and real value, and always, nonetheless, overshoots, because the overshoot is not a mistake by foolish people but a structural feature of how financial capital funds a revolution it cannot yet accurately price. Profitability at the centre does not immunise a build-out against overshoot at the margin; it merely relocates where the crash will fall.

So the recession now arriving is, in this frame, not the refutation of the AI revolution but potentially its installation-phase crash, the moment the theory says must come, when the financial enthusiasm is repriced against the technology's actual near-term returns and the overshoot is purged. And if that is what this is, then the useful questions are not the ones everyone is asking. Not "will it crash," because the frame says a crash is the normal mechanism by which a revolution transitions from frenzy to foundation. But rather: which of the players are laying down infrastructure that will still be valuable at sane prices after the repricing, the equivalent of the railways and the fibre, and which are the speculative froth that the crash exists to clear? Because that is what a crash actually does. It does not destroy the technology; it sorts the durable from the delusional, marks down the overbuilt infrastructure to the price at which it can finally be the foundation of the long deployment, and separates the companies that were building the future from the companies that were merely financing the belief in it.

I do not know whether the recession the oil shock is bringing will be the crash that sorts the AI build-out, or merely a scare before a larger one later. The timing, Perez would be the first to say, is never predictable. But I am fairly confident of the shape, because the shape has held across two centuries and every major technology: the frenzy, then the crash that everyone experiences as the end, and then, for the technologies that were real, the long deployment on the cheap foundation the frenzy left behind. The bulls are right that AI is real and will matter for decades. The bears are right that the current level of investment will not all be repaid and that a painful repricing is coming. Perez's frame is that both are correct, and that the crash the bears predict is the very mechanism that delivers the future the bulls believe in, which is why the question worth asking, as the recession arrives, is not whether the bubble bursts but which of the things being built now will still be standing, and useful, and cheap, on the far side of the burst.

The Ledger

The build-out meets its first recession; the framing above suggests the coming repricing is less a refutation than the installation-phase crash that every technological revolution passes through. The question shifts from "bubble or not" to "which of the overbuilt infrastructure survives the repricing to become the foundation."

Data centers & power

The overbuilding this paper has tracked looks, through the Perez lens, like classic installation-phase infrastructure: laid down in the frenzy, likely repriced in the crash, potentially the cheap foundation of the deployment to come. The data centres may outlast the valuations that built them.

Rates

The oil-shock inflation collides with recession risk, the textbook stagflationary bind; the Fed cannot cut into the inflation nor hike into the downturn, and the first corporate failures, an airline this fortnight, mark the shock's arrival in the real economy.

Real estate

US 30-year mortgage near 6.6 percent, held high by the war's embedded inflation even as growth weakens; the market braced between a recession that argues for lower rates and an inflation that forbids them.

India tech

India's frugal, capital-disciplined approach to technology, long noted in this paper, looks well-suited to a deployment era of cheap repriced infrastructure; a downturn that punishes overbuilding may reward exactly the constraint-first engineering culture Delhi has cultivated.

What we called wrong

Nothing to retract. The paper's three-year insistence that the AI build-out rested on a physical and financial foundation more fragile than the enthusiasm assumed is being tested now, in real time, by the first recession of the AI age.

The Back Page

The airline that ran out of sky

Spirit Airlines was not a beloved company; it was a byword for the cramped, fee-laden, no-frills flying that the budget traveller endured rather than enjoyed. But its failure this fortnight is worth a moment's attention beyond the schadenfreude, because it is the first clearly visible human cost of the war's economic shock reaching the ground, thousands of workers suddenly without jobs, routes to smaller cities suddenly without service, the abstraction of the oil price made concrete in the lives it touched. This is how a macroeconomic shock actually arrives: not as a number in a report but as an airline that cannot make the math work, a company that ran out of sky, and the people whose livelihoods went down with it. The war stopped a month ago, in another hemisphere, over the oil. And this fortnight, in the ordinary economy of ordinary countries, the bill began to be paid by people who never fired a shot and never set the price of a barrel, which is, as this paper has noted before and will again, the way the bills for wars are always, eventually, paid.