BlackRock’s Larry Fink Says Compute Could Become a New Asset Class — Is Crypto About to Hijack the AI Power Crisis?
BlackRock’s CEO says compute may become tradable like a commodity. Crypto bulls see a jackpot, but the real story may be energy, data centers and AI infrastructure.
When BlackRock CEO Larry Fink says the United States is short on power, compute, chips and memory, markets listen. When he adds that a future asset class may involve buying futures on compute, crypto accounts immediately translate it into one sentence: “trillions are coming into decentralized compute.”
That may be true one day. It may also be the kind of market slogan that turns a serious infrastructure problem into speculative rocket fuel.
The core idea is not absurd. Artificial intelligence is turning computing capacity into a strategic commodity. Data centers require land, electricity, chips, cooling, transmission infrastructure, financing and long-term contracts. In the same way oil, gas, wheat and electricity became tradable markets because supply and demand needed pricing mechanisms, compute could eventually become financialized. If AI demand keeps rising, companies may want to lock in future access to GPU clusters or specialized cloud capacity. Investors may want exposure. Traders may build derivatives. That is the world Fink is pointing toward.
But the phrase “compute futures” does not automatically mean small crypto tokens will capture the value. The biggest winners may be less glamorous: grid operators, data-center developers, chipmakers, cloud hyperscalers, power producers, cooling specialists, landowners, private-credit lenders and infrastructure funds. In fact, BlackRock’s own strategy has increasingly focused on infrastructure, energy and data centers. That tells us something. The institutional money may first flow into steel, concrete, electricity and contracts before it flows into decentralized compute coins.
Crypto still has a plausible argument. Decentralized physical infrastructure networks, GPU marketplaces and distributed compute protocols claim they can unlock idle hardware and compete with centralized cloud platforms. If compute becomes scarce and expensive, alternative networks could gain attention. They may serve startups priced out of hyperscaler contracts, AI labs seeking redundancy, or developers needing flexible inference capacity. In theory, tokenized incentives can coordinate supply.
The problem is quality. AI compute is not just “any computer somewhere.” Training frontier models requires reliable clusters, fast interconnects, predictable uptime, security, compliance and enormous capital expenditure. Inference can be more distributed, but even there, enterprise buyers need service-level guarantees. A viral crypto post can say “trillions will flow,” but institutional customers ask boring questions: who maintains the hardware, who guarantees uptime, who handles data privacy, who pays when the job fails?
There is also an energy angle. Fink’s comment about power may be more important than the compute-futures quote. The bottleneck for AI is increasingly electricity. The United States may have capital and demand, but connecting new data centers to the grid can take years. Local communities are beginning to ask why power-hungry AI facilities should receive priority. If compute becomes a tradable asset, electricity access may become the real underlying battlefield.
For investors, the critical distinction is between a theme and a trade. “AI compute shortage” is a theme. “Buy every decentralized compute token” is a trade. The first can be correct while the second still loses money. Many crypto sectors have experienced this pattern: a real macro trend appears, token narratives explode, valuations run ahead of adoption, and only a few networks survive.
The serious question is not whether compute becomes valuable. It already is. The question is who will own the pricing layer. Will hyperscalers create internal markets? Will Wall Street build standardized contracts? Will governments treat compute as strategic infrastructure? Will decentralized networks provide a credible alternative? Or will compute futures become another institutional product that uses the language of openness while concentrating power even further?
Fink’s comments should not be dismissed as hype. They are a signal that the AI boom is moving from software excitement into infrastructure scarcity. But investors should be careful with the translation. “Compute as an asset class” does not mean “every AI crypto token is underpriced.” It means the world is beginning to price intelligence production the way it prices energy, minerals and transport.
That may be bullish. It may also be dangerous. Once intelligence infrastructure becomes financialized, access to compute could become another arena where capital decides who gets to innovate.
The open question is simple: will compute markets democratize AI — or will they turn the future of intelligence into another commodity controlled by the biggest balance sheets?