BlackRock CEO Larry Fink says AI infrastructure needs trillions of Americans’ retirement savings

Larry Fink has a number in mind: $10 trillion. That is what the BlackRock CEO believes the United States needs to invest in AI infrastructure over the next decade, covering everything from data centers to power grids to the transmission lines connecting them. And he is not looking to Washington to write the check.

Fink’s argument, laid out at the Milken Institute Global Conference among other venues, is that the funding gap for America’s AI buildout is too large for public budgets to fill. The money, in his telling, needs to come from the people who already have it: ordinary Americans, through their retirement accounts, pension funds, and insurance policies.

Why pensions are suddenly the hottest topic in AI

BlackRock is well-positioned to broker that introduction. The firm entered 2026 managing approximately $14 trillion in assets, with more than half of that tied directly to retirement-linked accounts. It pulled in over $700 billion in net inflows during 2025 alone, giving it an enormous pool of capital to direct toward infrastructure projects.

The firm is not just theorizing. BlackRock has already established what it calls the AI Infrastructure Partnership, a vehicle designed to raise both equity and debt financing for AI-related construction. The partnership includes NVIDIA and Microsoft as collaborators, and BlackRock has floated a target of $30 billion on the equity side of the ledger.

Fink is also pushing a conceptual frontier: the idea of a new financial product built around “futures on compute.” The concept mirrors how energy markets work, where buyers and sellers can lock in future prices for electricity or natural gas. Applied to computing power, it would let companies hedge against fluctuating AI processing costs the way airlines hedge jet fuel.

Supply crunch, not a bubble

One of Fink’s more pointed arguments is directed at skeptics who see AI investment as the latest speculative frenzy. He contends the current environment is defined not by excess but by shortage. Power capacity is constrained. Computing resources are oversubscribed. The physical infrastructure required to run large-scale AI systems is simply not keeping pace with demand.

The national security dimension also runs through Fink’s case. Competition with China over AI capabilities has moved well beyond a talking point in corporate boardrooms. U.S. officials across both parties have framed semiconductor supply chains, data center capacity, and AI research leadership as strategic assets. Fink is threading his investment thesis through that needle, positioning private capital as a tool of national competitiveness rather than just a return-generating exercise.

What this means for institutional investors and the broader market

The emergence of compute futures as an asset class could be particularly significant for the financial industry. Energy futures transformed commodity markets and gave rise to entire trading desks, hedging strategies, and structured products. A parallel market for computing capacity would create analogous opportunities, and analogous risks, for firms willing to move early.

The risk for everyday retirement savers is less obvious but real. Infrastructure investments can be illiquid, meaning money tied up in a decade-long data center project cannot be easily redeployed if circumstances change. Fees on private infrastructure vehicles also tend to run higher than those on index funds.

The $10 trillion figure gives a sense of the scale Fink is describing. For context, the entire U.S. federal budget for a single year runs roughly in that neighborhood. BlackRock is betting it can be the institution that makes the coordination between financial firms and technology companies happen.