AI Debt Issuance Reaches $450 Billion as Bank of England Warns of Risks
Key Facts
Global AI-related debt issuance reached about $450 billion from the start of the year to early September, according to a Morgan Stanley estimate cited by the Bank of England in Financial Policy Committee minutes published on 30 September 2026. Issuance exceeded twice the total for the whole of 2025, amid committee warnings about the risks accompanying the expansion in financing. In the sterling bond market, issuance by AI-related cloud infrastructure hyperscalers accounted for 47% of corporate bond issuance from the start of the year to the preparation of the minutes. JPMorgan analysts estimated that sector-related capital expenditure expected to be financed through debt issuance would total about $4.1 trillion between 2026 and 2030. That estimate puts borrowing at the heart of funding for anticipated capital expenditure over that period.
The committee warned that rising indebtedness, opaque financing and circular arrangements could make risks harder to assess. It said these factors could amplify losses if AI-related expectations disappoint. It judged that a reassessment of the productivity gains expected from AI could affect sovereign debt markets, alongside valuations of the sector's own assets. The minutes reviewed the decline in AI and semiconductor stocks in July, which forced some investors to unwind leveraged positions. Nevertheless, market functioning remained "orderly", according to the minutes, and the disruption did not spread to core markets.
The expansion in sector-related borrowing had already featured in the bank's Financial Stability Report in December 2025. Using PitchBook data, bank staff estimated that new lending to AI companies in leveraged finance and private debt markets rose to 3 times its previous level between 2023 and 2024. It then rose again to nearly $100 billion from the start of 2025 to the preparation of the December report that year. In its July 2026 Financial Stability Report, the bank cited an Organisation for Economic Co-operation and Development estimate on the growing role of private credit in financing AI investment. According to the organisation, the share of this financing rose from 9% in 2024 to 34% in 2025.
In a more recent example of the sector's continuing search for financing, Reuters reported in October that SpaceX was in talks with banks and asset managers to raise $40 billion. The talks aim to finance purchases of AI chips from Nvidia, according to the agency. On the technology's use and governance, Bank of England Governor Andrew Bailey acknowledged its potential to support scientific discovery in a separate article published on the day the September minutes were released. He said AI could also boost productivity and economic prosperity. But he stressed the need for society to retain an effective ability to intervene as the technology develops.
Bailey called for rigorous testing of models "before and after deployment" to understand their behaviour, identify weaknesses and assess safeguards. He warned that advanced AI increases the scale and complexity of cyber threats. He linked the resilience of payment networks, banks and market infrastructure to the development of these models. At the same time, he stressed the limits of testing, saying: "Testing will not eliminate failures." He called for learning from incidents during development and use, as well as from near misses.
On the future of oversight, Bailey said model testing should not be seen as a substitute for future regulation. He added that testing was not a complete solution and that a more formal regulatory framework could emerge over time. On monetary policy, the Bank of England's Monetary Policy Committee held interest rates at 3.75% at its September meeting. The decision to hold was approved by a 6–3 vote, according to the monetary policy summary and meeting minutes. The next monetary policy decision is scheduled to be announced on 5 November 2026.