IMF Sounds Alarm: $1.4 Trillion in Circular AI Deals Threaten Market Crash

Three pillars of market turmoil

The International Monetary Fund's Global Financial Stability Report has named an AI-sector correction as one of three critical risks capable of sharply tightening financial conditions. The second threat is a spike in government bond yields, and the third is cyberattacks on the financial system.

According to the IMF, a correction could be triggered by weakening AI investment and falling profits. It could be made worse by rising borrowing, concentrated positions, and so-called circular deals in AI — structures in which investors loop risk back into promised returns.

The $1.4 trillion architecture: how the risk circulates

The IMF analyzed 14 publicly disclosed deals struck since 2023. In these, hyperscalers (the largest data center operators) channeled roughly $300 billion of their own capital into financially more vulnerable AI companies. At the same time, those same companies took on $650 billion in debt.

In return, hyperscalers signed long-term contracts entitling them to $1.4 trillion in future revenue — through guaranteed chip purchases and service agreements.

The critical point: roughly $320 billion of that debt remains on the balance sheets of the recipient companies themselves. Financing and obligations are concentrated among a handful of major players. The IMF called this phenomenon "wrong-way risk": investments lose value precisely when future revenues shrink. At the press briefing, IMF Assistant Director Jason Wu called it a "triple hit."

A borrowing spiral

Hyperscalers have become key borrowers in global financial markets. In 2024, their share of investment-grade debt issuance reached 10%. A year earlier it was under 1%. At the same time, their free cash flow is declining — AI infrastructure spending shows no signs of slowing.

Other AI companies, especially non-cloud operators, carry even heavier debt loads with weaker profits. This creates a risk of weak links in the chain. Law firms are already preparing scenarios for how AI data center debt could spiral into a crisis.

Investors themselves are also borrowing more. Combined hedge fund gross positions and margin loans have grown since 2024: they now equal 15% of advanced-economy GDP, up from 10% a year ago. Retail investors are using leveraged ETFs to bet on AI stocks, which now make up 28% of such funds, up from 16%.

In the event of a correction, advanced Asian markets would likely be hit first, given their close synchronization with global AI stocks. The eurozone and the UK are less exposed, but could still face stress through their growing portfolio holdings of U.S. equities.

Cyberattacks with no time to patch

On the cyber threat front, the situation is critical. State-sponsored hacking groups, ransomware gangs, and insiders have gained access to frontier AI models. These models show strong capabilities in simulating network attacks. In June, IMF Chief Economist Kristalina Georgieva warned that Mythos-class models could be turned against the financial system.

The pace of attacks is accelerating. In more than 70% of cases this year, software vendors had virtually no time to issue a patch before a vulnerability was exploited — a 20 percentage point jump from a year ago. By contrast, in 2021 vendors had roughly 300 days to develop a patch.

Banks in question, IMF recommendations

The IMF's stress test applied a scenario combining an energy price shock, higher inflation, rising bond yields, and an AI correction. The global banking system as a whole would remain resilient. However, banks controlling 22% of global sector assets would end up in a weak position, with capital ratios falling below 7% — down from 18% a year ago.

The IMF recommends that authorities curb excessive borrowing tied to AI bets: setting limits on margin loans, leveraged ETFs, and derivatives. In parallel, it calls for putting government debt on a sustainable path and strengthening cyber defenses.

Meanwhile, in its World Economic Outlook, the IMF raised its global growth forecast for 2026 to 3.1% (up from 3% in July). According to Georgieva, AI investment is keeping global growth at the 3% level.