IMF Says an AI Investment Pullback Could Cut U.S. Output 1.8% by 2028

The International Monetary Fund used its annual meetings in Bangkok this week to say something Wall Street doesn't like hearing: a lot of the world's growth is riding on AI spending that has yet to prove it pays for itself.

In its World Economic Outlook, released Tuesday, Oct. 13, the IMF forecast global growth of 3.1% in 2026 and 3.4% in 2027. That is the baseline. The warning sits in the downside scenario.

The scenario

If AI investment drops sharply and asset prices fall with it, the IMF estimates world GDP would end up 1.1% lower by 2028. U.S. output would take a bigger hit, 1.8%.

The companion Global Financial Stability Report lists an AI correction as one of three "salient risks," alongside rising government borrowing costs and a major cyber-attack. The fund called the trio "plausible near-term catalysts for a sharp tightening of financial conditions and market turmoil."

Managing director Kristalina Georgieva put it in plain terms at a Bretton Woods Committee event. "We now have an economy that is standing on one leg, the AI leg," she said. "What if it wobbles?"

How the money moves

The IMF's sharpest concern is not that companies are spending big. It is how they are paying for it.

The report says hyperscalers, the companies running the largest data center fleets, have strong earnings from their core businesses. But their free cash flows are "rapidly declining" as they try to meet AI investment needs, "leaving debt financing increasingly necessary."

Then there are the circular deals, where one firm funds another's AI project and gets the money back when the recipient buys its products. The IMF reviewed 14 such transactions. Since 2023 they involve about $300 billion in equity investments and $650 billion in debt, in exchange for $1.4 trillion in future revenue commitments.

Jason Wu, assistant director of the IMF's Global Markets Analysis Division, spelled out the chain. "If there are questions about the monetization of AI, the equity investments will deplete in value, the debt may be difficult to service, and the revenue commitments in the future may not materialize."

The report calls this "wrong-way risk": the investments lose value at exactly the moment the revenue tied to them is impaired.

Concentration

The so-called Magnificent Seven (Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla) now make up more than 35% of the S&P 500, up from 12% a decade ago. Their combined market value has risen more than 50% in the past two years.

The IMF also flagged "growing evidence" that a successful AI rollout "may not require investment on the massive scale being undertaken" by hyperscalers, and that the economic gains "may not be as large as expected."

Athanasios Vamvakidis, deputy director of the IMF's Monetary and Capital Markets Department, said the rest of the world's exposure to U.S. assets, particularly AI, is the highest it has ever been. That supports global growth through wealth effects today, he said, but a correction "could have broad implications." Asian markets would be the most exposed outside the U.S., the report says.

Chief economist Silvana Tenreyro added a trade angle. Trade in AI goods grew 29% from 2023 to 2025, more than triple the 8.3% growth in global trade. Strip it out and trade growth would have been 4.6%.

Outside voices

The IMF is not alone. Investment analyst Bret Jensen wrote on Seeking Alpha that the five biggest hyperscalers (Oracle, Microsoft, Meta, Amazon and Alphabet) carry more than $1.3 trillion in debt on their balance sheets and roughly $3 trillion in off-balance-sheet allocations. He argued that debt markets are signaling a substantial chance of default on Oracle's debt over the next five years, and that half the infrastructure is being built for OpenAI and Anthropic, which he called "unprofitable" and "bleeding cash." Those are Jensen's claims, not IMF findings.

Dean Baker of the Center for Economic and Policy Research argued in a Friday analysis that AI is holding up the U.S. economy and that there is little sign it will deliver the productivity gains needed to justify the spending.

On Tuesday, shares in semiconductor companies fell after Broadcom's chief executive compared many AI investors to casino gamblers. A report that Chinese memory chipmaker CXMT is making progress on a new chip architecture also hit South Korean tech stocks.

The other side of the ledger

Not everyone reads the numbers as a bubble. The IMF itself notes that the investors in these deals are "generally financially sound" and that hyperscalers have strong core earnings. A downside scenario is a stress test, not a forecast.

The spending is also tied to competition with China. Heather Carroll, chief revenue officer at Path Robotics and a member of the bipartisan National Security Commission on Robotics for Advanced Manufacturing, told Fox News Digital the U.S. leads in AI software but risks falling behind in deployment. "China has a national framework around robotics and AI," she said, while the U.S. has nothing similar.

Inflation and scale

The size of the buildout is what makes the exposure large. JPMorgan expects AI infrastructure spending of around $1 trillion this year. Goldman Sachs puts it at 1.9% of U.S. economic activity. Columbia economist Stijn Van Nieuwerburgh, in a Brookings paper, projects $10.3 trillion through 2032, averaging above 3.6% of GDP a year.

Economists cited by CNN worry that spending at that scale in an economy with low unemployment and strong consumer demand adds to inflation, on top of pressure from energy prices and tariffs.

The IMF's own baseline assumes the boom keeps funding itself. Its downside case depends on one thing: whether the $1.4 trillion in promised revenue shows up before the debt comes due.