IMF and Bank of England Warn of AI Correction, Flag $950B in Circular Deals

The two institutions whose job is to worry about financial stability have both put AI near the top of their list. On 30 September 2026 the Bank of England's Financial Policy Committee published the record of its 25 September meeting. It warned that a reassessment of AI could affect "not only AI-related asset valuations but also sovereign debt markets." Two weeks later, on 13 October, the IMF's October Global Financial Stability Report named a correction in AI-linked asset prices as a near-term risk. Its accompanying blog counted about $300 billion of equity and $650 billion of debt moving between AI firms in 14 deals since 2023. Three weeks after that, the Nasdaq was in a correction.

What happened

The Bank of England's committee spent most of its AI section on how the boom is paid for. "An increasing volume of AI-related investment is being financed through debt issuance," the record says. It cites Morgan Stanley's estimate that global AI-related debt issuance had reached about $450 billion by early September, more than double the total for 2025. It also cites JP Morgan's estimate of about $4.1 trillion of AI capex financed by debt over 2026 to 2030. Morgan Stanley expects private credit to fund $700 billion of data centre capex between 2026 and 2028. The committee flagged the "circular arrangements" in some AI financing, together with rising indebtedness and opacity. It warned that these could make risks harder to judge and make losses bigger if expectations disappoint.

The IMF's message, written by Athanasios Vamvakidis and Jason Wu, was similar. AI is holding up markets through data center and chip investment and strong earnings. Yet "faltering AI investment and profitability could challenge optimistic expectations," and a correction could set off wider deleveraging. The authors said hyperscalers are borrowing more and investors are using leverage to buy AI stocks. They also described a contagion channel inside "circular financing arrangements involving equity and debt flows between AI firms". Stress at the company receiving the money cuts its revenue forecasts and valuation, and that weakens the debt tied to the deal. The IMF's recommendation was blunt: "Leverage and liquidity mismatches in AI, private credit, and other nonbank financial sectors should be limited."

Private investors were saying the same in surveys. Bank of America's October Global Fund Manager Survey, taken 2 to 8 October from 161 participants managing $453 billion, found 37% overweight technology stocks, the highest since August 2020, investingLive reported. Long global semiconductors was again named the most crowded trade.

The numbers

Why it matters

Neither institution said AI is a bubble, and neither forecast a crash. What they did is describe the plumbing a crash would travel through. The Bank of England's point about sovereign debt is the one engineers tend to skip. Government growth and fiscal plans now partly assume AI productivity gains. If those gains get marked down, the damage goes beyond chip stocks to the bond markets that set everyone's borrowing costs. The IMF made the same link in the other direction: higher sovereign yields raise discount rates and push AI valuations down.

That is roughly what markets then did. After the US 10-year yield traded at its highest in more than two decades in early October, CNBC reported it was back below 5% by 22 October as investors sought cover from AI worries. On 5 November the Nasdaq fell 5.52% and closed in a correction.

What sceptics say

The reports include their own counterweights. The Bank of England noted that AI valuations, although they "fell sharply in July", remained "underpinned by strong expectations of future earnings", and those earnings are real. Nvidia reported $96.2 billion of quarterly revenue, up 106%, in its August release. The IMF's stress test was mostly about banks in general, and it found global banks resilient overall in its adverse scenario. It did flag that banks holding 22% of global banking assets would fall below regulatory capital limits, up from 18% a year ago. Regulators also have a habit of warning early. The Bank of England called equity valuations stretched in October 2025, and the Nasdaq went on to set new records in 2026.

My take, as opinion: the useful part of these reports is not the bubble verdict. It is that two official bodies now treat circular vendor financing as a stability issue in its own right. When the supplier is also the lender, the demand signal and the credit risk are the same bet.

What it means for engineering and DevOps teams

• Your cloud bill sits downstream of these bond markets. Hyperscalers are funding capex with debt, and the IMF and the Bank of England both flag higher sovereign yields as a risk. Expect less discounting and stricter commitment terms, and plan cost controls accordingly.
• Concentration risk is an architecture choice. If your AI features depend on one model provider that depends on one financing structure, you have the same correlated exposure the IMF is describing. A second provider behind an AI gateway is cheap insurance.
• Watch the boring indicators. CDS spreads on neoclouds, hyperscaler bond yields and private credit redemptions tell you about capacity pricing months before a price list changes.

FAQ

Did the IMF say AI is a bubble?

No. The October 2026 report and blog describe an AI correction as a plausible risk and explain how it could spread through leverage, circular financing and nonbank lenders. They do not forecast one or put a probability on it.

What is circular financing in AI?

It is when equity or debt flows from one AI company to another that then spends the money on the first company's products or services. Chip vendors funding customers who buy their chips is the classic case. The IMF counted about $950 billion of equity and debt in 14 such deals since 2023, tied to $1.4 trillion of future revenue commitments.

What did the Bank of England say about AI in 2026?

Its 30 September record warned that AI investment is increasingly financed by debt and that circular arrangements make risks harder to assess. It also said a reassessment of AI could hit sovereign debt markets as well as AI valuations.

Related: GPU Clouds in 2026: H100s at $2.59 an Hour, $104B of CoreWeave Backlog on $35B of Debt, and a Six-Year Bet on Silicon, The 2026 SaaS Selloff: IGV Fell 36%, Then Rose 45%. Revenue Grew 11% to 37% the Whole Time and FinOps for AI in 2026: 98% Manage Token Spend, 3 in 4 Can't Prove the Value, GPUs Run at 5%, and FOCUS Gets a Token Column in 1.5.