The warning itself dates back to the end of August. What is new is the response from OpenAI. George Osborne, the former UK chancellor who now oversees the AI company’s international business, has described the concerns raised by financial regulators as a “wake-up call”. The issue goes far beyond the familiar question of whether AI stocks are overvalued. Equity markets, hedge funds and debt financing are becoming increasingly intertwined with the AI boom. At the same time, powerful AI systems could accelerate cyberattacks against the financial sector. Regulators are therefore worried not simply about a crash — but about contagion.
Bailey is not predicting an imminent stock market crash
Andrew Bailey is not only Governor of the Bank of England. He also chairs the Financial Stability Board, or FSB, the international body that monitors risks to the global financial system on behalf of the G20.
In a letter to G20 finance ministers and central bank governors dated August 31, Bailey warned that markets remained vulnerable to a potentially disorderly correction that could spill across borders. Among the vulnerabilities he highlighted were elevated valuations, risks in sovereign bond markets and weaknesses in the rapidly expanding private-credit sector.
This is not a forecast that a crash is about to happen. The concern is the combination of risks. Several parts of the financial system appear vulnerable at the same time. If stresses emerge simultaneously, they could reinforce one another. And that is where the AI boom becomes relevant.
Around half of the S&P 500 is now linked to AI
The Bank of England had already devoted a separate section of its July Financial Stability Report to the issue. According to the Bank, companies classified as AI-related now account for roughly 50 percent of the S&P 500’s market capitalization. At the end of 2022, the figure was around 25 percent.
The calculation does not refer only to pure-play AI companies. It includes a defined group of businesses with substantial exposure to the development and commercialization of artificial intelligence.
That turns a potential valuation problem into a concentration problem. If a relatively small number of major AI-related companies disappoint investors’ expectations for future earnings, their share-price reaction can move a much larger part of the overall market than it could only a few years ago.
The Bank of England also notes that some US equity valuation measures have returned to levels seen only rarely in the past. Even after the 30 stocks most closely associated with AI are excluded, valuations remain elevated.
AI is therefore not the sole reason US equities look expensive. But it has increased market concentration.
Hedge funds are taking larger leveraged positions
The picture becomes more complicated when investors use borrowed money. According to the Bank of England, hedge funds’ global prime-brokerage exposures have risen by around 40 percent within a year, reaching record levels. At the same time, hedge-fund positions have become increasingly concentrated in individual sectors, including semiconductors.
When markets rise, leverage amplifies returns. When prices fall sharply, the same mechanism works in reverse. Hedge funds can be forced to sell positions rapidly in order to meet margin calls or reduce leverage. And they do not necessarily sell only the securities that caused the original loss.
Many funds operate simultaneously in equity and government bond markets. A sharp fall in AI stocks could therefore trigger sales in entirely different asset classes. Those are precisely the kinds of connections financial regulators watch because they can turn an ordinary repricing into a disorderly market move.
The AI boom is increasingly being financed with debt
Only a few years ago, the main financial risk surrounding artificial intelligence was that investors might simply be paying too much for technology stocks.
The financing structure itself is now changing. According to the Bank of England, major AI companies reached a turning point in 2025: the scale of investment required for data centers and related infrastructure increasingly exceeded the amount that could comfortably be funded from operating cash flow alone.
External financing consequently accelerated sharply in the first half of 2026.
The five AI hyperscalers examined by the Bank — Meta, Alphabet, Amazon, Microsoft and Oracle — issued more bonds in the first half of 2026 than they did during the whole of 2025. At the end of 2025, these companies accounted for only three percent of the outstanding US investment-grade corporate bond market. By early May, however, they already represented more than 15 percent of new issuance during the year.
And the scale of investment could rise further. Market expectations cited by the Bank suggest that capital spending by major AI companies in 2028 has been revised upward within only a few months — from less than $600 billion to more than $1 trillion.
When an equity story becomes a credit story
Debt is not automatically a problem for highly profitable companies. Meta, Alphabet and Microsoft have strong balance sheets and high credit ratings. The Bank of England explicitly notes that credit markets have so far absorbed the additional issuance without difficulty. There is currently little evidence that other borrowers are being crowded out.
The warning concerns what could happen next. The more AI investment is financed through bonds, private credit, structured financing and off-balance-sheet arrangements, the more the risks move beyond the stock market into other parts of the financial system. There are also what regulators describe as capital loops.
Technology companies sometimes invest in AI businesses that then use that capital to purchase computing power, chips or infrastructure from the same companies or their partners. The Bank of England sees the potential for self-reinforcing financing structures. As long as demand and profits grow as expected, those arrangements can work extremely well.
The problem begins if one central assumption changes — for example, how quickly companies adopt AI, how much they are willing to pay for it or how profitable the technology ultimately proves to be. A repricing would then no longer affect shareholders alone.
And now an entirely different AI risk is emerging
Bailey’s second warning is easily confused with the first. The FSB is not saying that autonomous AI agents are about to trigger a stock market crash.
Instead, it identifies cyber risk as the most immediate threat to the financial system arising from advanced frontier AI.
New models increasingly possess autonomous capabilities and can solve complex problems with less human intervention. According to the FSB, that could fundamentally change the speed, scale and cost of cyberattacks.
An attack that once required considerable expertise, manpower and time could increasingly be automated and carried out in parallel against large numbers of targets.
For banks, exchanges or payment infrastructure, the consequences could extend beyond the technical disruption itself. If confidence in critical systems is damaged, a cyber incident can become a financial-stability event.
Now OpenAI is responding
That is the part of the warning George Osborne has now addressed publicly. The former UK chancellor is currently responsible for OpenAI’s international activities. In a contribution published in the Financial Times, he described Bailey’s warning as a »wake-up call«.
One of the central concerns, Osborne argues, is that AI is becoming capable of discovering vulnerabilities faster than companies can fix them. The answer, he says, cannot simply be to deploy AI to find more weaknesses. Systems must also help organizations prioritize vulnerabilities, repair them and then verify whether the fixes have actually worked.
Osborne is calling for closer cooperation between governments, financial institutions, cybersecurity companies and the leading AI laboratories. The fact that an OpenAI executive is publicly engaging with the warning gives the issue new relevance.
The real danger is not one single AI bubble
Perhaps the most important message from financial regulators is that the nature of the debate is changing. For a long time, the central question was:
Are Nvidia, OpenAI and other AI companies overvalued? Increasingly, the question is becoming:
What else is now tied to those valuations?
Equity indices have become more concentrated around AI-related companies. Hedge funds are using more leverage. Large technology groups are raising substantial amounts of capital to finance data centers. Private lenders and structured-finance providers are increasingly involved in the build-out. At the same time, the very AI systems driving this investment boom are becoming a new factor in the cybersecurity of the financial system itself.
None of this means that a crash is inevitable. The risk lies in the connections.
Financial crises rarely begin simply because one asset became somewhat too expensive. They become dangerous when falling prices force investors to reduce debt, sell collateral and expose vulnerabilities in places where few people expected them. The AI boom has now become large enough for financial regulators to start watching those connections closely.
That may be the most important change of all:
Artificial intelligence is no longer just a stock market story. It is becoming a financial-stability story.
SK
