Larry Fink, CEO of BlackRock, told the Milken Institute that surging demand for computing power could give rise to a new asset class — futures contracts on raw compute capacity. Fink compared the idea to energy and agricultural commodities, which already trade on futures markets. He also warned that the United States lacks enough chips, memory, and power to handle projected artificial intelligence workloads.
Comparing Compute to Corn and Crude
Fink drew a direct line between computing and traditional physical commodities. Just as farmers lock in corn prices and airlines hedge jet fuel, he argued, hyperscalers and AI firms may soon hedge compute costs. “Raw computing power is becoming a resource as essential as oil or wheat,” he said. The comparison isn't casual — energy and agricultural futures are deeply liquid markets, and Fink sees similar potential for compute. But a key difference remains: no standard unit exists yet for compute capacity across different hardware generations and AI workloads.
No Bubble, Just a Supply Crunch
Fink rejected the idea that AI investment is a bubble. He said demand still outpaces supply across the entire stack — from chips to data-center power. That imbalance, he predicted, will lead to a capital shortage in AI infrastructure. “We’re not overbuilding,” he told the audience. “We’re under-resourced.” His view contrasts with some market skeptics who worry about overinvestment. Fink’s message is that the shortage isn't temporary; it's structural.
BlackRock’s Hyperscaler Partnership
BlackRock is preparing to unveil a partnership with an unnamed hyperscaler to push deeper into AI infrastructure. The deal would move the world's largest asset manager beyond financing and into direct stakes in physical assets like data centers and power plants. Fink declined to name the partner ahead of the formal announcement, but the move signals a shift in how BlackRock views AI — not just as an investment theme, but as an operational bet. The firm manages $13.9 trillion in assets.
The Hurdle of a Standard Unit
Whether exchanges actually list compute futures depends on one thing: a standard unit of computing power. Right now a megawatt-hour of electricity is a fixed measure, and a bushel of corn is a bushel. But compute power varies by chip generation, workload type, and even cooling efficiency. Without a standardized contract, liquidity won't follow. Fink’s framing suggests BlackRock sees a workable benchmark arriving sooner than the market currently prices. That timeline remains an open question.
As for the hyperscaler deal, Fink kept the name under wraps. The formal announcement is expected in the coming weeks. In the meantime, the industry is left to figure out how to turn gigaflops into a tradeable future.
