A correction in AI-related stocks and debt could set off market turmoil, the International Monetary Fund said on Tuesday. It named the AI risk as one of three that could sharply tighten financial conditions. The other two are a sharp rise in government bond yields and cyber threats.
The warning came in the IMF’s Global Financial Stability Report at the annual meetings in Bangkok. Weaker or less durable AI investment and profits could trigger the correction, the Fund said. Rising borrowing, concentrated exposures and circular AI deals could make it worse.
The IMF reviewed 14 publicly disclosed deals since 2023. In them, investors led by hyperscalers put about $300bn of equity into financially weaker AI firms, alongside $650bn in debt. In return, they secured $1.4tn in future revenue commitments through services or chip orders.
About $320bn of that debt sits with the firms receiving the money, the report said. A handful of them hold most of the financing and revenue promises.
The IMF called it “wrong-way risk”: “Equity and debt investments in the firms depreciate in value precisely when future revenue streams tied to them are impaired.”
IMF assistant director Jason Wu called it a “triple whammy” at a press briefing, CNBC reported.
More borrowing
Hyperscalers, the biggest data centre operators, now issue 10% of the world’s investment-grade bonds. In 2024, it was less than 1%. Their free cash flow is falling fast as they keep spending on AI, the IMF said. That makes debt increasingly necessary.
Other AI firms, such as neocloud operators, carry more debt and are less profitable. The report said this raises the possibility of weak links. A law firm recently mapped how AI data centre debt could unravel.
Investors are borrowing too. Hedge funds’ gross equity positions plus margin loans have grown since 2024. They now equal about 15% of advanced economies’ GDP, up from 10%. Retail investors are using geared exchange-traded funds, which magnify returns, to bet on AI stocks. Single stocks now make up 28% of these funds, up from 16%.
Advanced Asian stock markets, which move most closely with global AI shares, would likely suffer the most in a correction. The euro area and the UK are less sensitive. Stress could still reach them through their sharply higher holdings of US shares.
Less time to patch
On cyber threats, the IMF said state-sponsored groups, ransomware gangs and insiders now have access to frontier AI models. Those models have shown strong capabilities in network-attack simulations. In June, IMF chief Kristalina Georgieva warned that Mythos-class AI models could be turned against the financial system.
Attackers are also moving faster. In more than 70% of cases this year, software providers had almost no time to patch a flaw before it was exploited. That is 20 percentage points more than a year ago. In 2021, providers typically had close to 300 days, the IMF said in a blog post.
Weak banks
The IMF’s 2026 bank stress test combined an energy price shock, higher inflation and bond yields, and an AI-led market correction. Banks would remain broadly resilient, the Fund said. But banks holding 22% of global banking assets would end up weak, with core capital ratios below 7%. A year ago, that share was 18%.
The Fund called on authorities to contain excessive borrowing behind AI bets, including limits on margin loans, geared exchange-traded products and derivatives. It also urged governments to put public debt on a sustainable path and strengthen cyber defences.
In its World Economic Outlook, also published on Tuesday, the IMF raised its 2026 global growth forecast to 3.1%, from 3% in July. On Monday, Georgieva said AI investment is keeping global growth near 3%.