Wall Street's most persistent tech bull is back with a characteristically contrarian take: the AI selloff that has wiped trillions from the Nasdaq since August is over, and the next leg up begins now. Tom Lee, Fundstrat's co-founder and managing partner, made the call Monday, citing a thesis that deserves more scrutiny than his headline prediction.
Lee's argument hinges on a distinction that has been underappreciated in the recent panic. The slowdown in AI spending, he contends, is not a demand problem but a supply problem. Hyperscalers cannot build data centers fast enough; Nvidia cannot ship GPUs fast enough; utilities cannot provision power fast enough. This is not the dotcom bust, where eyeball-chasing companies discovered that clicks do not equal revenue. This is a boom constrained by physics and permitting.
The supply-side story
The numbers lend some support. Microsoft, Amazon, and Google have all flagged capacity constraints in recent earnings calls. Nvidia's backlog extends well into 2027. The AI infrastructure buildout is proceeding at a pace limited by transformer manufacturing, not by corporate enthusiasm. Lee's interpretation: the stocks sold off because growth decelerated, but growth decelerated because the industry hit a ceiling, not a wall. When supply catches up, the ramp resumes.
This is a plausible reading, though it elides some inconvenient questions. Enterprise adoption of generative AI has been slower than the hype suggested. Many pilot projects remain pilots. The productivity gains that justified trillion-dollar valuations have yet to show up in aggregate economic data. Supply constraints are real, but so is the possibility that demand was always more speculative than the market admitted.
The crypto tell
Lee's second argument is more novel. He points to cryptocurrency's muted response to the Binance hack — the largest exchange breach in history — as evidence that risk appetite is not dead, merely resting. Bitcoin dipped but did not crater. Altcoins wobbled but stabilized. The sellers, Lee suggests, are exhausted. If crypto can absorb a $4 billion hack without capitulating, perhaps tech can absorb a valuation reset without spiraling.
It is a clever cross-asset read, though one that assumes crypto and AI equities share the same marginal buyer. They increasingly do — the same macro tourists who piled into both in 2024 have been fleeing both in 2026 — but the correlation is not destiny. Crypto's resilience could reflect idiosyncratic factors (the approaching Bitcoin halving, stablecoin inflows from emerging markets) rather than a broad risk-on signal.
Our take
Tom Lee has been wrong before, spectacularly and publicly, and he has also been right when consensus mocked him. The value of his call is not the prediction itself but the framework: supply constraints versus demand collapse, cross-asset risk signals versus single-sector narratives. If you believe AI infrastructure is a generational buildout temporarily gated by physical bottlenecks, his bullishness is defensible. If you believe the productivity case for generative AI remains unproven at enterprise scale, the selloff may have further to run. The honest answer is that both things can be true, and the market is still deciding which matters more.
