Comparable doesn’t mean identical
Renowned investors such as Michael Burry – famous for having anticipated the subprime crisis – are betting on a fall in the tech sector; surprising and even controversial comments from a number of tech bosses, not least those of OpenAI and Nvidia; and the word “bubble” is spreading across the financial press, accompanied by a question mark. Unquestionably, the optimism accompanying the growth of Artificial Intelligence (AI) in recent months is under pressure. And worries about a possible bubble against the background of a major technology innovation naturally raise the parallel with the internet bubble.
This analogy is regularly refuted by some market players who explain that the two situations are very different and conclude that the current environment is not remotely bubble-like. The reality appears to be more nuanced. One of the main arguments put forward is that current valuations are much more sustainable than in 2000, that they are less irrational and relate to sound and profitable groups. This justification appears relevant. With the exception of Palantir, which is trading on an estimated 2027 PER of 130x, valuations in the technology sector, whilst high, are far from the exuberance of those at the time of the internet bubble or the Japanese bubble of 1990. As a comparison, Microsoft today trades on a PER of 30x next year’s estimated earnings, whereas this ratio was over 60x in 1999.
But whilst valuations seem far less absurd, the weight of the technology sector is way more important today than in 2000. At that time, the cumulative weight of the ten largest companies in the technology, media and telecommunications (TMT) sector represented 20% of the S&P 500. Today they account for almost double that (39%). This comes in parallel with the very high concentration of US stock market indices – the ten largest companies by market capitalisation represent 42% of the S&P 500 versus 27% in 1999, as practically all “top ten” members are technology stocks. So, any bubble on the markets is more of a concentration bubble than a valuation bubble.
Another argument regularly posited to dispel the bubble theory is the modest number of IPOs, particularly in comparison to the frenetic activity at the end of the nineties. In reality, this is more a function of developments in financing. During the internet bubble, IPOs, i.e. equity markets, were the preferred source of new funds. Today, the favourite sources of financing are different, with a greater focus on debt. Firstly, through increased use of bond issues. In the US, the technology sector this year represents one third of investment grade new issues[1], compared to 15% usually. Secondly, from the use of private markets, whether through fundraising or the use of private debt. The private market platform Forge Global estimates that the funds raised by 19 AI companies represented three quarters of all capital raised on private markets in 2025.
Lastly, financing between companies within the same AI ecosystem, in particular Nvidia, which has invested directly in several of its clients, such as OpenAI, Nscale and CoreWeave, to enable them… to buy Nvidia chips! This practice raises fears of “inbreeding” within AI investments. We should also mention the emergence of financial engineering techniques such as that which enabled Meta to remove around USD 30 billion of debt from its balance sheet by parking it within an SPV[2] financed by Blue Owl Capital and Pimco. For the most pessimistic, such techniques are a reminder of certain precedents of the era prior to the subprime crisis. So, whilst 2000 saw a bubble in equity and capital markets, the current situation looks more like a bubble in debt and private markets.
If comparable doesn’t mean identical, this applies in both directions. Against the backdrop of better corporate profitability and abundant liquidity in the financial system, it is, undoubtedly, difficult to compare current equity market valuations to those at the time of the internet bubble. In contrast, just because we have not seen rampant irrational behaviour or a surge in IPOs is not enough to dispel suspicions of a bubble. Furthermore, the situation should be seen in the context of a rapid rise in debt, including in the weakest segment in balance-sheet terms of the AI environment. Investors should bear in mind the words of Mark Twain: “History doesn’t repeat itself, but it often rhymes”.