Could AI Create the Next Financial Crisis? The Bank of England’s Warning
IT Club — Powered by Altitude AI (opens in a new tab)
Keep up with IT Club
Add IT Club as a preferred source in Google Search.
The Bank of England’s September 2026 Financial Policy Committee record links rapid AI-related borrowing to a wider set of financial vulnerabilities. Here is what the warning says, what the headline figures mean and how SMEs can test AI investments before committing.
The Bank of England has not said an AI-led financial crisis is inevitable. Its Financial Policy Committee (FPC) said on 30 September 2026 that the chance of several financial vulnerabilities crystallising together had risen. One concern was the rapid increase in debt linked to AI investment.
That distinction matters. AI may produce valuable products and productivity gains, while some companies and investors may still be taking on more risk than future revenues can support. The FPC’s warning is about the financing and connections around the technology—not a claim that AI itself is destined to fail.
What did the Bank actually warn about?
The committee described a wider risk picture involving sovereign borrowing, risky asset valuations and credit markets. It said the financial system had remained resilient so far, but exposures were becoming more interconnected. The rapid growth of AI-related borrowing broadens the number of investors exposed to whether AI companies meet expectations.
The record cited outside estimates to show the scale. Morgan Stanley estimated that AI-related businesses had raised more than US$450 billion in debt by early September 2026, more than twice the total for 2025. The FPC also cited a JPMorgan estimate of about US$4.1 trillion in debt-financed AI infrastructure and equipment investment between 2026 and 2030. These are analyst estimates and forecasts, not a measured Bank of England total or a guarantee that all the spending will happen.
Why does AI need so much capital?
Large AI services depend on expensive physical infrastructure: specialist processors, servers, data-centre buildings, electricity, cooling, network connections and backup capacity. Companies may spend heavily before it is clear how many customers will pay, how quickly hardware will become outdated or whether operating costs can be recovered through subscriptions and usage fees.
Borrowing can make sense when expected returns are strong and predictable. The risk rises when firms depend on continued investor funding, optimistic future demand or complex deals between companies that buy from and invest in one another. If revenue disappoints, loans still need to be repaid. If a large supplier or borrower cuts spending, the effect can spread to lenders, contractors, landlords and customers.
This does not mean that AI has no genuine value. It means that the commercial case for each investment should be separated from the most optimistic forecasts about the overall market. A sound use case can make sense even if another company’s data-centre expansion or valuation does not.
A practical AI investment test for SMEs
Small businesses do not need to forecast the global AI economy to make a sensible purchasing decision. They do need to know what they are buying and what happens if the promised gains do not arrive.
- 1Name the problem and baseline. Record the current time, cost, error rate or service level before introducing a tool.
- 2Calculate the full cost. Include licences, implementation, staff training, integration, security review and ongoing support—not just the monthly fee.
- 3Run a bounded trial. Use a defined workflow and success measure, then compare results with the baseline before expanding.
- 4Check supplier dependence. Ask where your data is processed, what happens to it, how you can export it and what service continuity exists if the supplier changes pricing or closes.
- 5Set an exit point. Agree who can stop the service, how records will be retrieved and what manual process will keep work moving.
The Bank’s warning is a reason to ask better questions, not to abandon useful technology. Treat AI as an investment with costs, risks and a measurable return. If a proposal has no clear business problem or exit plan, the excitement around the market is not a substitute for a business case.
Sources and further reading
Bank of England: September 2026 FPC record →
Reuters: Bank warns risks from AI and debt are growing →
IT & Security News: Bank of England warns of AI debt risks →
Plain-English Takeaway
AI should solve a business problem, not become an expensive solution looking for one. Set the expected benefit, full cost, risk limits and exit plan before committing.
Related Articles
AI Needs Data Centres. But Who Should Pay for Them?
AI may feel like software, but it depends on very physical infrastructure. Who should pay for the electricity, water and grid investment behind the AI boom?
Read articleIf AI use really grows 100,000-fold, who's paying for all that computing?
Huawei forecasts that global annual AI token consumption could grow 100,000-fold by 2035, with agent traffic over 90%. It is a forecast—not a guaranteed future—and it gives SMEs a useful way to think about AI economics.
Read articleCould AI Charge You More Than the Next Customer?
The reporting concerns restaurant-specific price recommendations—not evidence that McDonald’s charges individual customers different prices.
Read articleNeed help putting this into practice?
IT Club helps you understand the technology. If you need implementation, support or consultancy, the teams behind IT Club can help.
Altitude IT (opens in a new tab) — IT support, cyber security, Microsoft 365 and technology operations.
Altitude AI (opens in a new tab) — AI discovery, automation, governance and implementation.