The market surrounding artificial intelligence technologies has reached a scale that leaves the legendary dot-com bubble of the late 1990s in the dust. Based on my estimates, derived from a detailed analysis of market data, the parallels between these two eras are not just obvious — the current AI boom surpasses its predecessor across a number of critical metrics.
Key Metrics: Infrastructure and Capital
The first and most telling indicator is the volume of capital investment in infrastructure. During the dot-com era (1999-2001), total infrastructure spending amounted to around $500 billion. Today, in the AI era (2025-2026), this figure has soared to $5.8 trillion. That's more than a tenfold increase.
The size of public offerings (IPOs) also shows a colossal gap. The largest IPO of the dot-com era raised roughly $4 billion. In the current cycle, the record stands at $86 billion. This points to an unprecedented appetite among investors for AI companies, even when many of them lack profitability.
Of particular note is the valuation of unprofitable companies. The largest "unprofitable" company of the dot-com era was worth about $100 billion. Its modern counterpart is valued at nearly $1 trillion — $965 billion. This is a massive premium for future, as-yet-unrealized revenues.
Signs of Overheating: Lessons from the Past
Analyzing the current situation, I see alarming signs characteristic of an overheated market. Unprofitable companies are going public at the peak of the hype. The example of SpaceX, which turned negative just a month after the largest IPO in history, is a stark confirmation of this.
Another dangerous signal is the customer financing scheme. Nvidia, much like Lucent and Nortel in the late 1990s, has launched a program to finance its own clients. History shows that this practice led to some of the largest bankruptcies in U.S. history.
Depth and Layering of the Bubble
The main difference — and the main danger — of the current bubble is its multi-layered nature. The dot-com bubble was largely "paper-based" and consisted of stocks. Today's bubble includes private loans, project bonds, and even insurance company funds. All this capital is flowing into data center infrastructure.
For example, Blue Owl froze investor withdrawals while simultaneously financing 80% of Meta's (banned in Russia) $250 billion campus in Louisiana. Oracle's rating was downgraded to nearly "junk" status: half of its order book is tied to OpenAI. Anthropic has $90 billion in capacity lease obligations with zero profit before its IPO. This is a fragile structure that could collapse at the slightest slowdown.
My Verdict
AI is a real and powerful technology, much like the internet was in 1999. The question is not about the potential of the technology itself, but about its valuation. Do companies justify the sums investors are paying for them? The answer will come from public reports in the coming quarters. However, given the scale of accumulated debt and the complexity of financial structures, the consequences of a potential downturn could be far more severe than in 2000. The market should brace for high volatility and, possibly, a painful correction.