BlackRock Doesn't Bet on Who Wins AI. It Bets on Whoever Wins Needing Chips.
February 22, 2026 by Asif Waliuddin
BlackRock's AI-focused ETF just increased its overweight on AI infrastructure names -- chips, data centers, networking -- during a broader AI-related selloff.
The rotation: away from pure-play model vendors, toward picks-and-shovels plays.
The reasoning: more predictable cash flows.
Translation: the largest asset manager on earth looked at the AI market and concluded they are not confident enough in any specific model company's long-term position to hold concentrated exposure.
Here's what most people are missing:
-- BlackRock is not selling AI. They are selling model vendor risk. The distinction matters. They believe AI infrastructure will be needed regardless of which model company wins. They do not believe they can pick the winner.
-- This is happening during a selloff. When stocks dip after capex announcements and BlackRock rotates INTO infrastructure rather than out, they are buying the dip on chips and selling the uncertainty on models. That is a conviction trade, not a panic trade.
-- The pattern rhymes with every major technology cycle. In the cloud era, the picks-and-shovels trade was networking equipment and data center REITs. In the mobile era, it was chip designers and tower companies. The common thread: when the winners are uncertain, bet on the inputs that all winners need.
-- Nvidia reduced its OpenAI commitment by 70% the same week. The chip company and the largest asset manager made the same read: the infrastructure layer is more predictable than the model layer.
When the smart money rotates from model vendors to infrastructure, the message is not "AI is failing." The message is: "We believe in AI. We do not believe we know who wins."
That is a more honest position than most AI investment theses.
Follow for more AI Hype vs Reality takes.