For the first time in the history of financial regulation, the International Monetary Fund officially classified the AI boom as a full-fledged bubble, capable of rivaling the dot-com crash of 2000-2001 in terms of scale and consequences. This is not an abstract warning, but a statement of systemic risk delivered at IMF meetings in October 2025. The financial community has exhausted the tools to prevent collapse.
The scale exceeds historical precedents. In recent months, the capitalization of the top 10 technology companies has reached $24.6 trillion – an absolute maximum. The entire dot-com bubble created a similar amount in 5.5 years. Today’s growth is almost three times faster, and this unprecedented rate of capital formation is the first sign of severe overheating.
All revaluation metrics have reached extreme levels. The P/S rose from 4.85 (2019) to 9.53, an increase of 96%. The P/E rose to 38.1 (was 26.9, +42%). The multiplier shows that with current earnings, it will take 38 years to recoup the share price. Capitalization to free cash flow: 54.1 vs. 25.3 six years ago (+114%). According to Bank of America, 60 percent of bear market indicators have already been activated — the historic collapse threshold of 70 percent. All 20 tracked metrics are in the revaluation zone, nine of them exceeding the peak values of the dot-com bubble.
Concentration has reached critical levels. Top-10companies make up 33.8% of the capitalization of S&P 500—a maximum from 1972year. The five giants account for more than a third of the index’s total growth. Without these companies, operating cash flow remains virtually zero. This dependence creates the conditions for a “domino effect” in any crisis.
The fundamental gap between expectations and reality. OpenAI is valued at $1trillion with revenue of just $13billion. The MIT study found that 95% of the 300 AI projects surveyed remain unprofitable, despite investments in the tens of billions. AI-investment accounted for 2% of U.S. GDP in 2025 year — without the AI segment, American business would have stagnated. The market is living on expectations of exponential productivity growth, and real returns do not provide a valuation trajectory.
One of the key drivers of growth is infrastructure projects worth hundreds of billions of dollars. Companies like OpenAI, Broadcom, AMD, and Oracle are contracting with each other to form a vicious circle of investment: each finances the other’s projects, and capital actually moves in circles. Such circular financing gives the impression of large investments and market legitimacy, but these transactions are often not backed by the stable cash flow necessary to service obligations.
Investment is focused on less regulated lending. A significant part (about 58%) of AI capital is financed through non-banking organizations without proper regulatory oversight. Approximately 50% of US banks have investments in such structures that exceed their Tier 1 capital – the “cushion” for survival in a crisis. Essentially, banks have invested in high-risk assets, and now the potential losses are greater than reserves. In a bankruptcy crisis, they will go one after another.
Temporary factors mask underlying weakness. Advanced purchases, absorption of costs at the expense of margin, incomplete transfer of tariffs to prices only create the illusion of stability in 2025 year. All factors are temporary buffers that are exhausted. By the end of the year, the full inflationary effect of tariffs will put pressure on corporate profits and consumer demand. However, we will discuss this in more detail in the next post.
The main conclusion: the AI bubble of 2025 year is formed record quickly and concentrated, and the financial stability of the system is significantly weaker than with dotcoms. Connectivity of banks and non-bank structures, lack of monetary maneuvering and weak fiscal margin make the consequences of the collapse especially severe.