To be or not to be in a bubble

• Enguerrand Artaz
  Strategist
  La Financière de l’Échiquier (LFDE)

The market context is conducive to such jitters: since its highs in early October, the Nasdaq 100 has fallen by 7 per cent, whilst over the same period, the US semiconductor index (SOX) has plummeted by 19 per cent¹. At the same time, the setbacks faced by the cloud start-up CoreWeave have put pressure on the credit market. But is it still appropriate to keep talking about a ‘bubble’?

Inevitably, this term brings to mind the year 2000 and the dot-com bubble. However, the situation in the equity markets is hardly comparable. At an average of 25.6% per annum over the last three years, the Nasdaq 100 is certainly performing strongly, but it is far from the exuberance of the late 1990s, when, between 1995 and 2000, it rose by an average of 58.6%² per annum. The same logic applies in terms of valuation. With the 12-month forward price-to-earnings ratio³ (P/E ratio) expected to be slightly above 20 times, the Nasdaq’s valuation is trading below its 20-year average. One could rightly argue that this moderate valuation is due to very high earnings expectations. Nevertheless, if we reduce these earnings forecasts by 30 per cent, the Nasdaq’s P/E ratio rises to around 29 times, which is close to what it was in 2021 at the peak of the post-Covid rebound. This is certainly a high level, particularly given the current interest rate environment, but it remains a far cry from the 60 to 65 times seen in 1999.

Whilst equity markets are undeniably expensive, perhaps a little complacent and far from immune to a sharp fall should the AI boom come to a halt, it makes little sense to talk of a bubble given the current valuations and performance trends. It is even possible to argue that the markets are reasonably carrying out the necessary adjustment. Take the example of semiconductors. Whilst questions are beginning to arise as to whether the growth peak of the semiconductor cycle is not nearly upon us, the valuation of the global semiconductor index has already fallen to 0.9 times that of global equities, which is close to the lowest levels seen in the last 20 years. In the credit market, CDSs⁴ for the most indebted companies, such as CoreWeave, have already widened significantly. This shows that bondholders are not turning a blind eye to the balance-sheet fragility of certain AI players and are not waiting for an imminent risk of default to materialise before demanding higher yields.

In our view, therefore, it is not the state of the markets that investors should be concerned about at present. If there is a bubble in AI, it is only very partially visible in the listed markets. This potential bubble relates to capital expenditure, particularly on data centres, and, even more so, to the financing arrangements for this infrastructure. The risk stems from a financing chain that has become massive, complex and circular. Computing power contracts, signed by players still reliant on fundraising, enable the financing of GPUs and data centres, the profitability of which in turn depends on sustained high demand. If the monetisation of AI were to prove disappointing, the creditworthiness of clients, the value of contracts and that of the assets provided as collateral could all deteriorate simultaneously. Lenders would then tighten their terms, forcing the most indebted players to scale back their investments or sell off their capacity at a loss. This would accelerate the fall in prices and recovery values. The losses could then be passed on to private debt funds, insurers, securitisation vehicles and, ultimately, the real economy.

Far more than in stock market multiples, it is here – in the financing – that we believe the risk of an AI bubble currently lies.

__

Disclaimers: This data and these opinions from LFDE, as well as the sectors and securities mentioned, are provided for information purposes only and, as such, do not constitute an offer to buy or sell any security, nor do they constitute investment advice or financial analysis. Past performance is not indicative of future results and is not consistent over time.

__

¹Performance as at 22 October 2026
²Performance excluding reinvested dividends
³Company valuation ratio
⁴Credit Default Swap: insurance against the risk of default by a debt issuer.