AI Bubble Becomes Credit Investors' Top Fear for the First Time

Financial market sentiment is slow to change. When professional investors who manage trillions of dollars in debt instruments change their biggest fears in unison, it deserves close attention. That is exactly what happened in February 2026: according to a regular Bank of America survey of clients of its credit division, the "artificial intelligence bubble" took first place among the key market risks for the first time ever. Not geopolitics, not inflation, not recession, but AI.

To grasp the scale of this shift, it helps to recall the context. Throughout 2023 and 2024, the tech sector went through euphoria comparable to the dot-com boom. Nvidia's market capitalization grew several-fold, and Microsoft, Google, Amazon and Meta raised capital spending on data centers and AI infrastructure at a pace analysts called unprecedented. In 2025 alone, the combined investment of the "big four" cloud providers in AI infrastructure exceeded $250 billion by various estimates. Meanwhile, the question of return on that investment stayed open: revenue from AI products was growing, but clearly not fast enough to justify such outlays.

Credit investors are not venture optimists or retail traders buying stocks on hype. They run bond portfolios and assess companies' ability to service debt over five, ten, even twenty years. When this group of market participants starts calling the AI bubble its main worry, it is not abstract philosophizing. It is about concrete credit spreads, borrowing costs and the availability of capital for the entire tech sector. In essence, the debt market is sending a signal: we are no longer sure the enormous AI spending will pay off, and we are ready to price in that risk.

There are plenty of reasons for the pessimism. First, generative AI revenue growth at most large companies began to slow toward the end of 2025, not because the technology is bad, but because corporate customers have been slower to adopt it than expected. Putting AI into business processes takes more than an API subscription: it requires rebuilding infrastructure, training staff and resolving data security questions. Second, competition among AI model providers has driven inference prices down sharply, which is good for consumers but squeezes margins. Third, the emergence of efficient open models has called into question the very business model of spending billions to train proprietary systems.

It is important to understand that credit investors' concerns are not a verdict on the technology. The internet bubble of the early 2000s burst, but the internet did not go away, and it ultimately transformed the economy just as the boldest visionaries had promised. The problem was not the technology but valuations and the pace of monetization. The analogy with AI suggests itself: artificial intelligence really is changing entire industries, but the path from lab demos to sustainable corporate revenue has turned out to be longer and more winding than markets assumed during the euphoria of 2023-2024.

The practical consequences of this shift in sentiment could be quite significant. If credit markets begin to systematically reprice tech sector risk, borrowing costs will rise across the whole chain, from the largest cloud providers financing data center construction to AI startups raising venture debt. This will not necessarily lead to a crash, but it could significantly slow the pace of infrastructure investment. For AI startups the signal is even more worrying: when even credit investors doubt the durability of the trend, raising money for the next round will become considerably harder.

The Bank of America survey records not a collapse but a turning point in perception. The market is moving from a phase of unconditional faith in AI to a phase of demanding scrutiny: show the revenue, show the return on investment, show that your customers actually pay rather than merely experiment. This is a painful but ultimately healthy process.

Companies with real products and proven value for customers will get through this period of doubt. Those who built businesses on presentations and promises will be in a much tougher position. The era of free money for anything with the letters "AI" in its name is coming to an end, and that may be the best thing that could have happened to the industry.

*Meta has been designated an extremist organization and banned in Russia.