Citrini, an investor who drew attention for his bullish stance on AI hardware, has since adjusted his positions. In an interview, he said he had reduced his semiconductor exposure, held very few semiconductor stocks, and had some short positions. That does not mean he is bearish on chips across the board: he remains positive on Nvidia, but believes more attractive opportunities may lie in other AI applications.

His biggest concern is the changing way AI infrastructure is financed. In the past, tech giants largely built computing capacity with cash generated by their businesses, giving spending a fundamental basis even when market sentiment fluctuated. Now, a growing share of new investment relies on equity and debt financing. Citrini believes that even if AI continues to advance and usage keeps growing, investors could still pay too much for capacity. If sentiment turns, the flow of new funding could also dry up quickly.

As for specific trades, he said he currently has a “long Meta, short Google” position. He is watching the risk posed by changing search habits, as well as the pressure Google faces from talent departures. At the same time, he acknowledged that the eventual impact on its search business remains uncertain.

On AI drug discovery, he pointed to a constraint that is harder to scale quickly than computing power: a shortage of laboratory monkeys needed for preclinical testing. Computer simulations may eventually reduce the need for animal testing, but reliable models still require experimental data today. Because breeding monkeys takes time, supply cannot increase as quickly as AI-driven research.