Could AI-linked debt become a financial-stability risk? The Bank of England widens its warning
The Financial Policy Committee says AI investment is exposing more investors and funding markets to a repricing. Its official record quantifies the build-out, but the underlying analyst estimates are not a stress test or forecast of losses.
The short version
- What happened
- The Bank of England says AI-related debt is spreading exposure to AI outcomes across investors and funding markets, while high valuations leave room for a sharper repricing.
- Who it affects
- Workers and small businesses could feel a financial shock through tighter credit or delayed investment even if they hold no AI shares directly.
- What remains unproven
- The record reports the committee's assessment and selected market estimates; it is not a peer-reviewed causal study or an issuer-level audit.
Keep in mind
The record reports the committee's assessment and selected market estimates; it is not a peer-reviewed causal study or an issuer-level audit.
The underlying analyst methodologies for the $450 billion, $4.1 trillion and $700 billion estimates are not reproduced in the public record.
Orderly markets and a 2% countercyclical buffer do not eliminate risk, but the report does not establish that an AI-related loss event is imminent.
Source analysis · 2 linked sources. Primary analysed source date: 30 September 2026. Earliest linked supporting source: 7 July 2026.
Read the full report · 5 min The evidence, methods, implications and limitations
What changed in the Bank's assessment
The Bank of England's Financial Policy Committee published the record of its 25 September meeting on 30 September. The committee says the likelihood of interconnected financial vulnerabilities crystallising has risen since July. Its concern is broader than a fall in technology shares: debt-funded AI investment is linking model development, data-centre construction, corporate credit, private markets and the expectations built into wider economic forecasts.
The record identifies two AI channels. First, a disappointment in expected earnings, adoption or productivity could reprice AI-related shares and credit. Second, more capable frontier models may increase cyber and operational risks for financial firms. The committee says test-environment incidents involving autonomous models taking unexpected actions add to the need for containment, monitoring, patching and governance. It does not claim that an AI-driven financial crisis is under way, and it says markets have remained orderly so far.[1][2]
The scale—and the denominator behind the figures
The FPC cites a Morgan Stanley estimate of roughly $450 billion in global AI-related debt issuance by early September 2026, more than double the total for all of 2025. It also cites a JP Morgan estimate that $4.1 trillion of AI-related capital expenditure could be financed through debt between 2026 and 2030, plus a Morgan Stanley estimate that private credit could finance $700 billion of data-centre investment from 2026 to 2028. In sterling corporate-bond markets, AI-related issuers accounted for 47% of issuance so far this year, according to the record.
Those numbers are material, but their construction matters. The published FPC record does not reproduce the analysts' issuer lists, classification rules, currencies, maturities or treatment of project finance and refinancing. The $450 billion denominator is issuance classified as AI-related, not the total debt of AI companies, the amount already impaired or an estimate of taxpayer exposure. The $4.1 trillion and $700 billion figures are forward-looking market estimates, not commitments. They should be used as scale indicators rather than precise loss forecasts.[1]
What the warning means for firms and investors
For lenders and asset managers, the practical question is whether apparently diversified exposures depend on the same assumptions: sustained demand for computing, reliable access to energy, rapid model progress and eventual productivity gains. The Bank says opacity and sometimes circular financing arrangements can make that dependence harder to see and could amplify losses if expectations disappoint. Private-credit structures may add another layer because public information about valuations, covenants and counterparties is often thinner than in listed markets.
For financial firms, the operational warning is separate from the investment case. A bank could be financially exposed to AI infrastructure while also relying on advanced models or suppliers in cyber defence, customer service and internal operations. The committee urges firms to use regulator and National Cyber Security Centre guidance and sector information-sharing groups. A sensible board-level exercise would map credit, equity, supplier and cyber dependencies together, then test what happens if valuations fall while an operational incident interrupts a critical service.[1]
What it means for people—and what would change the assessment
Households are not being told to expect an imminent banking shock. The committee kept the UK countercyclical capital buffer at its neutral 2% setting and judged that UK banks remained appropriately capitalised and liquid. The public-interest issue is transmission: if losses caused lenders to pull back, businesses could face tighter credit and workers could feel the consequences through investment and employment. Pension savers may also hold AI-linked companies and credit indirectly through funds without seeing the concentration clearly.
This assessment would become more concerning if audited issuer-level data showed rising leverage, weak covenants, concentrated short-term refinancing, correlated private-credit exposures or stress tests that produced spillovers into core markets. It would become less concerning if financing were transparent, maturities were well spread, cash flows covered debt under conservative demand assumptions and cyber exercises showed that essential services could continue through a frontier-model incident. The September record is therefore an early system-level warning with concrete market indicators—not evidence that the AI build-out has already become a financial crisis.[1]