If you took one editorial board’s advice this week and went for a relaxing stroll instead of checking your 401(k), you missed the only story that matters. What happened to AI stocks this week was not the “rude reminder” some commentators called it. It was a warning shot across the bow of an economy that has loaded up on a genuinely transformative technology with debt, opacity, and the sort of loose-money enthusiasm that has ended badly every single time it has been tried. Hype about a new technology is a dangerous collective delusion that almost never pays off. Call this week what it was: a catastrophic reckoning, and investors are right to demand returns on capital immediately.
Many tech shares took a dive, sometimes by large percentages. The NASDAQ and the S&P 500 went with them. That’s a fact. What’s not a fact is the soothing story built on top of it — that encouraging earnings reports from some big AI-related companies justify the optimism. In some cases the good earnings news arrested the decline in a company’s stock price. But that is beside the larger point, which is that stock prices are forward-looking, not backward-looking, and investors are now grappling with a question the reassuring voices keep dodging: not whether AI is real, but whether anyone is going to get paid for it.
AI will have a devastating effect on the global economy. It remains impossible to say what those effects might be, or who will profit from them, and how. Estimates of global AI-related investments regularly top $1 trillion per year and are almost certainly undercounts after including the companies adopting AI in their own businesses. That number is supposed to reassure you. It should not. One recent academic study calculated that American data-center hyperscalers alone would need to earn revenue equal to roughly 9% of gross domestic product per year by 2032 to break even on their investments. Nine percent of GDP. From one corner of one industry. By 2032. A number like that should frighten anyone doing the arithmetic, because it implies AI would have to deliver returns no technology in history has ever approached — and to concentrate control of the productive economy in a handful of firms in order to do it. The bet is being made with the entire productive economy as collateral.
The selloff this week is what markets look like when participants begin to suspect the gap is wider than they had been willing to admit. The reassuring framing — that this is healthy discipline, that markets were bound to “remind” investors about returns on capital eventually — mistakes the symptom for the cure. The cure would have been a much earlier, much less generous period of price discovery. What we got instead was years of heady valuations, increasingly debt-financed, increasingly routed through opaque private-credit markets, and increasingly exposed to exactly the kind of reappraisal some commentators are now assuring us is unlikely.
That view found “unforgiving market discipline on the way down” to be a “not-so-secret ingredient in the special sauce.” It is hard to think of a more upside-down reading. The special sauce has been the patience of long-duration capital — institutional investors and entrepreneurs, the legal system and the market structure — that allowed the AI buildout to be financed at scale in the first place. Discipline on the way down is what you want in a commodity boom. In a technology revolution you want something closer to discipline on the way up: sober underwriting, transparent financing, balance sheets that don’t depend on hidden leverage. What America’s ability to raise and allocate private capital has actually produced here is a bubble of historic proportions, marketed to the world as the envy of civilization and run in reality like a casino propped up by easy money and self-deception.
And on monetary policy the comfort is upside down too. Loose money didn’t merely encourage the credit mania — it manufactured it. But the answer to a bubble built on liquidity is not to drain the punchbowl mid-plunge. Less rum at the wrong moment in a technology revolution is the precise kind of premature tightening that turns a healthy price correction into a credit event, especially with so much of this buildout financed through opaque private-credit markets where the next bloodletting is being assembled in silence. Federal Reserve Chairman Kevin Warsh’s reported intention to “pour less rum into the punchbowl” is being greeted as a welcome development. It is not. The least-worst posture at the Fed right now is the opposite of what is being applauded: hold liquidity in place, watch the parts of the system where the risk is concentrated, and be ready to act rather than preen. Credit will turn a technological reappraisal into a financial panic, full stop.
All of this poses an existential challenge for politicians and policy makers, who have profited from the equity ride up and will be held accountable on the way down. The regulatory half of the challenge is to recognize that the corners of the financial system where catastrophic risks lurk are not corners at all but the entire structure. The goal here is not to shield everyone everywhere from any loss. It is to admit that markets cannot operate effectively in the correction already underway — and to admit that public bailouts are already baked into the trajectory. Hey, we can stop pretending.
AI as a technology may not be a winner. It may be the most overhyped failure since the railroads, and the cost of finding out will be borne by the workers, retirees, and taxpayers who took the ride up. Whether any of the companies currently developing it will be winners is the only question that matters, and the editorial case made this week ducks it. The market is already beginning to give the answer.
So no — now is not the time for a walk. Reconsider your entire AI exposure instead of pretending a weekend stroll will make it go away. Read the 10-Qs. Read the private-credit disclosures, where they exist. Read the academic study that puts the 9% number on the page. Then decide for yourself whether the gap between invested capital and earned revenue is closing as fast as the AI bulls need it to. If the answer is no — and the answer this week is no — then the so-called reminder is the opening chapter of something the reassuring voices do not want to name.
