The AI bubble has overshadowed the dot-com era: unprecedented scale and hidden risks
The current frenzy around artificial intelligence has significantly surpassed the famous dot-com bubble of the late 1990s by a number of key indicators. The parallels between the two eras are becoming increasingly obvious, but the scale of the current "overheating" inspires much greater concern.
History teaches us: even revolutionary technologies do not guarantee success for everyone who bets on them. In the dot-com era, the internet was a real force, but the vast majority of companies from that time went bankrupt without ever building sustainable business models. Today, we are witnessing something similar, but on a much larger scale.
Numbers that speak for themselves
The key parallel between the two bubbles is the scale of capital investment. If in the dot-com era (1999–2001) total spending on infrastructure amounted to about $500 billion, then in the AI era (2025–2026) this figure has reached a staggering $5.8 trillion. An increase of more than 11 times.
Even more striking is the contrast in the size of public offerings (IPOs). The largest IPO of the dot-com era raised about $4 billion. Today's record holder in the AI sector raised $86 billion — a difference of 21.5 times.
Particular attention is drawn to the valuation of unprofitable companies. The most expensive "unprofitable" asset of the dot-com era was worth about $100 billion. Today's counterpart is valued by the market at $965 billion — almost ten times more.
Signs of overheating: déjà vu with new risks
The market is showing classic signs of a bubble. Unprofitable companies are going public at the peak of the hype. For example, after the largest IPO in history, one of the leading AI companies went into the red within a month.
A worrying signal is also the scheme of lending to buyers. The largest manufacturer of chips for AI has launched a financing program for its own clients. In the late 1990s, Lucent and Nortel did the same, which ended in some of the largest bankruptcies in US history.
Why the current bubble is more dangerous
The main danger of the current bubble is that it is "hidden under a layer of stocks." The dot-com bubble consisted mainly of securities. Today's includes private credit, project bonds, and insurance company money. All these funds are flowing into data center infrastructure. This multi-layered structure makes the situation potentially more severe.
Vulnerabilities among key players are already visible. One of the large data center operators has frozen investor withdrawals, while simultaneously financing 80% of Meta's (banned in Russia) campus worth $250 billion. Oracle's rating has been downgraded to nearly "junk": half of its order book is tied to OpenAI. Anthropic has $90 billion in capacity lease obligations with zero profit before its IPO.
My analysis: AI is a real technology, like the internet in 1999. But the question is not about the technology, but about the price. Do the companies justify the amounts investors are paying for them? The answer will come from public reports, and judging by current metrics, the correction could be painful. In the cryptocurrency market, we are used to volatility, but the scale of the intertwining of traditional and private capital in the AI sector creates systemic risks that could also affect our markets. Keep an eye on the reports — they will be the trigger.