Correction in AI stocks: what is the best strategy to avoid a bubble? - February 9, 2026
The stocks most sensitive to the AI boom theme are experiencing a correction and high volatility, with significant disparities over the past three months. This nervousness is somewhat reminiscent of the year 2000, which reinforces the attractiveness of ultra-diversified strategies that performed well at the time.
Much more than the extreme volatility of precious metal prices, it is the theme of artificial intelligence (AI) that continues to raise questions for equity investors. The recent decline in AI stocks is part of a correction that began at the end of October 2025. Since that high point, the Bloomberg index of stocks most sensitive to AI has lost 12.5% (as of 5 February 2026), with a very turbulent ride. For the moment, this volatility is not spreading: the S&P 500 has remained relatively stable, and the globally weighted universe has gained nearly 5% (Chart 1).
Within the multifaceted theme of the AI boom, a considerable divergence has emerged between hardware stocks, which benefit directly from investments in data centres and electrical equipment, and software stocks, which finance this boom in order to develop solutions (Chart 2).
This dichotomy is not so surprising: the increased profitability of hardware companies is based on concrete orders for the construction of data centres, while that of software companies remains hypothetical. Some of the latter – even though they remain highly profitable – are suffering in the short term from a decline in profitability linked to the considerable capital expenditure they have incurred. Finally, the competitive landscape is changing rapidly, raising doubts about the future profitability of certain long-standing players. This echoes the questions that arose during the telecoms and internet boom in 2000, and even during the railway boom in the 19th century.
In terms of valuation, however, AI stocks have already undergone a significant correction. After reaching 30x at the end of October 2025, the P/E ratio of the Bloomberg AI index fell back to just over 22x, which is significantly lower than the valuations seen for the ‘TMT’ theme during the 2000 bubble (over 40x). Compared to the globally equal-weighted universe, the relative valuation of the AI theme has even returned to its pre-ChatGPT launch level (Chart 3).
The arguments in favour of a market scenario modelled on that of 2000 are therefore not all convergent. But if history were to repeat itself, equally weighted strategies – which are ultra-diversified in terms of geography and sector by design – could come out on top, as has already been the case for the past three months. While the Nasdaq lost 40% in 2000, an equally weighted portfolio of global equities from developed countries (large and mid caps) gained +6%, significantly outperforming the MSCI World (-13%). The equity exposure of Dorval AM's Global Macro strategies is based on this principle.
