The artificial intelligence (AI) market is demonstrating a scale that far surpasses the infamous dot-com bubble of the late 1990s. An analysis by expert Hedgie reveals striking parallels between these two eras, but with one critical difference: today's bubble is significantly larger and potentially more dangerous.
Key Indicators: Numbers That Speak for Themselves
A comparison of infrastructure costs vividly illustrates the difference in scale. During the peak of the dot-com era (1999–2001), total infrastructure spending amounted to about $500 billion. In the AI era (2025–2026), this figure has soared to $5.8 trillion — an increase of more than 11 times.
A similar picture emerges in the realm of initial public offerings (IPOs). The largest IPO during the dot-com era raised about $4 billion. Today's record holder stands at $86 billion. The valuation of unprofitable companies has also undergone colossal changes: the most expensive unprofitable company of the dot-com era was worth about $100 billion, while its modern counterpart is valued at $965 billion.
Signs of Overheating: History Repeating Itself
Expert Hedgie notes alarming signs that were already observed in the late 1990s. Unprofitable companies are going public at the peak of the hype, and SpaceX, for example, went into the red just a month after one of the largest IPOs in history.
Particularly concerning is the buyer financing scheme. Nvidia has launched a program to finance its own customers — exactly like Lucent and Nortel did in the late 1990s. This, as is well known, ended in some of the largest bankruptcies in U.S. history.
Why Today's Bubble Is More Dangerous
The main danger, according to the analyst, lies in the fact that the AI bubble is "hidden beneath a layer of stocks." Unlike the dot-com bubble, which mainly consisted of securities, today's bubble includes private credit, project bonds, and insurance company money. All these funds are flowing into data center infrastructure.
Signs of vulnerability are already visible. Blue Owl has frozen investor withdrawals while simultaneously financing 80% of Meta's (banned in Russia as an extremist organization) $250 billion campus in Louisiana. Oracle's rating has been downgraded to nearly "junk" status: half of its order portfolio is tied to OpenAI. Anthropic has $90 billion in capacity lease obligations with zero profit before its IPO.
My analysis: The parallels drawn with the dot-com bubble are not just a historical reference, but a serious warning for the market. AI is a real technology, just like the internet was in 1999. But the question is not about the technology, but about its valuation. Do companies justify the sums investors are paying for them? The answer, as before, will come from public reports. And if they disappoint, the consequences for the market could be catastrophic, especially given the multi-layered debt structure supporting this bubble.