The Bank of England’s Financial Policy Committee cited a Morgan Stanley estimate that global AI-related debt issuance had reached around $450 billion by early September, more than double the total for 2025, in minutes published on 30 September, ANI reported.
Key points
- JPMorgan analysts estimated that debt could finance around $4.1 trillion of AI-related capital expenditure between 2026 and 2030.
- Morgan Stanley analysts estimated that private credit would fund around $700 billion of data centre capital expenditure from 2026 to 2028.
- The committee warned that leverage and complex financing arrangements could magnify losses if expectations for AI investment fall short.
JPMorgan’s $4.1 trillion debt estimate
The committee also cited a JPMorgan estimate that debt issuance could finance around $4.1 trillion in AI-related capital expenditure between 2026 and 2030, ANI reported. The estimate covers spending across that period. Morgan Stanley’s figure for debt issued by early September measures borrowing already raised, giving the committee both a current exposure and a projection for its possible growth.
AI borrowing from January to early September surpassed the $333 billion of UK government gilts planned for issuance in 2026, the Guardian reported. Hedge funds, asset managers and private credit firms are among the investors exposed to AI companies through the debt, at a time when those businesses have yet to turn a profit, according to the Guardian.
Morgan Stanley analysts estimated that private credit would finance around $700 billion of data centre capital expenditure between 2026 and 2028, according to the committee’s account reported by ANI. The committee identified rising leverage, limited transparency and what it called “circular arrangements” in AI financing as factors that could make risks harder to assess and increase losses if investment expectations are not met.
Borrowing costs are another part of the financing picture. AI Affairs previously reported that bond buyers were demanding wider spreads from AI-linked issuers as hyperscaler debt supply grew. It also reported that a SoftBank junk bond drew more than $20 billion in demand for an OpenAI investment.
July’s AI share losses and leverage
The committee’s warning followed a sharp fall in AI-related and semiconductor shares in July. Volatility forced some leveraged investors to close positions, and the ensuing deleveraging and portfolio changes intensified equity-market moves, Quartz reported. Market trading remained orderly despite significant losses for some firms with concentrated holdings, and the episode did not spill into core markets.
The committee said hedge fund leverage remained elevated and AI valuations were still supported by strong expectations of future earnings. Doubts about the pace of AI development or adoption could prompt a sharper repricing, it warned. It also said the risk of a correction reaching core markets persisted, despite the orderly trading during July’s decline.
Expectations for AI-driven productivity gains have become important to growth prospects and fiscal outlooks, the committee said. A reassessment could therefore affect sovereign debt markets as well as AI-linked assets. Separately, the Bank for International Settlements warned that a surge in AI investment spending could end in a stock-market crash and recession, Quartz reported.
Bailey calls for AI model testing
Bank of England governor Andrew Bailey addressed a different route by which AI could affect finance: the operation of increasingly capable models. In a separate opinion piece, he argued that society should retain the ability to intervene in AI systems. He called rigorous testing of new models a “sensible starting point” for understanding their behaviour and identifying where intervention could work, the Guardian reported.
The committee also pointed to models released during the third quarter of 2026 that showed greater ability to complete complex tasks without human direction and to find and exploit software vulnerabilities in testing environments. It linked those advances to cyber and operational risks. The committee maintained the UK countercyclical capital buffer at 2%, saying banks remained appropriately capitalised and held high levels of liquidity, Quartz reported.
