Is artificial intelligence transformational? Are investors getting caught up in a bubble? Seasoned investor, Howard Marks tackles this question in a thoughtful memo that acknowledges both the transformative potential of AI and the warning signs of excess.
February 17, 2026
The Core Tension
Marks distinguishes between two different bubble questions: Are companies investing wisely in AI, and are investors behaving rationally? While he can’t judge corporate behavior, he focuses on investor psychology—and finds unmistakable signs of "irrational exuberance." AI stocks have driven 75% of S&P 500 gains despite representing a small fraction of companies. Nvidia alone has appreciated dramatically since its initial IPO, a rise that cannot be ignored.
Why This Might Be Different
Unlike the dot-com bubble, AI already has real products generating exploding revenues. Anthropic doubled its revenues year-over-year; coding services are growing 10-100x annually. Major companies like Microsoft and Alphabet trade at reasonable P/E ratios—actually lower than during the late-1990s tech bubble. Existing demand is genuine and massive.
But The Risks Are Real
Marks highlights enormous uncertainties: Which companies will win? Will AI become commoditized, and when will it actually become “profitable”? ChatGPT and Gemini lose money per query. Startups are raising $10-50 billion with no product or revenue. Concerning "circular deals" between tech companies—reminiscent of the telecom bust—create questions about illusory profits. Debt is financing speculative assets with 30-year bonds backing infrastructure of unknown lifespan.
The Historical Pattern
Bubbles around transformative technologies (railroads, electricity, the internet) do accelerate progress—but destroy investor wealth in the process. Marks notes that past revolutionary technologies generated bubbles, crashed, then rebuilt at lower costs. The question isn’t whether AI will be transformative; it’s whether current valuations justify the enormous unknowns.
The Bottom Line
Marks advocates a middle path: maintain a moderate, selective position with prudence. Don’t go all-in without acknowledging ruin risk, but don’t stay all-out and miss a generational opportunity. As one observer noted, the enthusiasm around AI may be excessive, but whether it’s irrational remains unknowable—bubbles are best identified in retrospect.