The governor of the Bank of England has issued a warning to investors about the return on investment (ROI) of artificial intelligence (AI).
In an interview on , Andy Bailey said: “There is a very large amount of investment going into the AI sector. That’s natural because it’s a major area of growth.”
However, while he believes AI needs financing given its potential to be a major contribution to growth to the UK and world economy, he warned that its contribution may not be as great as financial markets expect and it could take a lot longer to reap those financial benefits.
His comments come just days after Reuters revealed a leaked prospectus for Anthropic’s initial public offering (IPO), valuing the AI firm at $2tn. Bloomberg also interviewed OpenAI CEO, Sam Altman, who spoke about the need to get AI safety right before being held accountable to investors, as and when the company eventually goes public.
Responding to a question from Bloomberg about the company’s IPO, Altman spoke about the increased capabilities of frontier AI models and the need to implement AI safety: “While we are going through this period of adjusting to this new level of capability and the new safety requirements, we just want to get our feet under us, make sure we understand how to operate in this new way and be able to make some of these decisions without the pressure of being a newly public company.”
The Bank of England’s Record of the Financial Policy Committee (FPC) meeting of 25 September 2026 reported that developments in AI could affect a wide range of investors and funding markets. The committee noted that an increasing volume of AI-related investment is being financed through debt issuance.
“The increasing indebtedness of AI firms combined with opacity and, at times, ‘circular arrangements’ that can be associated with this financing, can complicate the assessment of risks and could amplify losses if expectations disappoint,” the FPC warned.
The committee also warned that growth prospects and fiscal outlooks are partially dependant on the expectation that AI development and adoption will generate significant productivity gains. The FPC said: “A reassessment of those expectations could therefore affect not only AI-related asset valuations but also sovereign debt markets.”
In addition, the FPC discussed how a narrowing of the current capability gap between open-weight and closed-weight AI model capabilities would make more advanced capabilities widely accessible. This presents a potential AI safety risk as open models could be exploited by cyber criminals, but it also means that advanced AI capabilities are also available to everyone without the premium that is generally associated with the advanced closed models from frontier AI firms.
Clearly, if advanced AI capabilities are available as open models, technology buyers may choose these potentially cheaper models over the most advanced models on offer from the likes of Anthropic and OpenAI. In addition, open models tend to require less powerful hardware, which means the token cost of an open model together with the AI acceleration hardware cost are both significantly less than closed models.
Investors in AI firms may have to wait more than five years before they see any measurable returns, according to analyst Gartner. The analyst firm projected that AI capital investment in aggregate would not achieve a return on invested capital (ROIC) of 12% – the investment hurdle rate – until 2032 at the earliest.
From an IT buyer’s and CIO strategy perspective, there is a very real risk that some of the businesses they have bought AI products from may not be around by then. There is also a risk that AI inference and machine learning costs will rise significantly as publicly listed AI companies look to increase shareholder value and fund continued investments in expensive AI acceleration hardware.