2025: A YEAR OF THREE MARKET DECLINES
2025 was a year marked by extreme price volatility. We weathered three significant declines, and at each one the market commentariat declared the AI trade was over. In moments like these, the pressure to ‘do something’ — to trade around the noise, cut winners, or move to perceived safety — is enormous. It would have been easy to capitulate.
We did not.
Our inaction was not stubbornness, but the philosophy we presented to you at the start of the year: we do not watch the waves on the surface of the Atlantic; we watch the Gulf Stream beneath them. In our Q1 2025 letter, written during the height of the China-tariff scare, we argued that the fundamentals of our thesis were unchanged. We stressed in particular Jevons’ paradox — the economic reality that when compute costs fall, demand does not slow; it explodes. We held to that all year.
In 2025, the largest risks came from rational thinking. Software companies rationally protected their margins. Analysts rationally depreciated down hardware acquisition costs. Both decisions were wrong, because the economic rules of the AI era have turned upside down — at least partly.