“Let the Market Decide”: Sam Altman Rejects Government Bailouts for the AI Revolution

OpenAI CEO Sam Altman clarified that the company will scale computing aggressively but will not take government guarantees for private data centres. He argued governments could build and own national AI compute reserves for strategic uses, while the market should decide which private firms succeed. Altman outlined financing plans, a $1.4 trillion infrastructure commitment, and why the company is investing now. 

Sam Altman’s recent post on X landed like a policy brief wrapped in a founder’s plea: OpenAI will invest at scale to build the computing backbone it believes the world will need—but it will not ask taxpayers to underwrite private data centre expansion.

Here is a detailed post: 

I would like to clarify a few things.

First, the obvious one: we do not have or want government guarantees for OpenAI datacenters. We believe that governments should not pick winners or losers, and that taxpayers should not bail out companies that make bad business decisions or…

— Sam Altman (@sama) November 6, 2025

Why Altman rejects guarantees — and what he does support

“I would like to clarify a few things,” Altman wrote, then left little room for ambiguity. “We do not have or want government guarantees for OpenAI datacenters. We believe that governments should not pick winners or losers, and that taxpayers should not bail out companies that make bad business decisions or otherwise lose in the market.”

That line is the thesis of Altman’s message: commercial scale is necessary, but accountability must remain with private actors. Where governments can and should step in, he suggests, is in building public-purpose compute — national reserves of computing power that serve strategic objectives and whose upside flows to the state, not private balance sheets.

Market discipline, public-purpose compute

Altman separates two distinct policy choices. On private datacentres: no guarantees, no bailouts. On strategic capacity: governments could build and own compute and offer lower-cost capital to support that infrastructure. “We can imagine a world where governments decide to offtake a lot of computing power and get to decide how to use it,” he wrote, framing state-owned compute as a public good rather than a corporate subsidy.

He also flagged a narrower exception where public finance has a role: semiconductors. Altman acknowledged discussions around loan guarantees tied to U.S. semiconductor fabs — not to prop up datacentres, he emphasised, but to rebuild chip supply chains for national security and industrial policy. “This is of course, different from governments guaranteeing private-benefit datacenter buildouts,” he said.

How OpenAI intends to pay for scale 

Altman answered the predictable follow-up: how will OpenAI pay for its aggressive buildout? His financial framing was stark: OpenAI expects to finish the year above a $20 billion annualised run rate and is planning roughly $1.4 trillion in commitments over the next eight years to expand computing infrastructure. Revenue levers include enterprise contracts, consumer devices, robotics, speculative scientific-use cases, and selling compute capacity as an “AI cloud”.

Crucially, Altman positions this as a market bet, not a plea for state underwriting. He acknowledged that the company may raise more equity or debt, but the overarching message was confidence that demand will match the investment if OpenAI executes.

Too big to fail? No — but insurance for catastrophic risk

Altman pushed back on the “too big to fail” narrative. “If we screw up and can’t fix it, we should fail, and other companies will continue on doing good work and servicing customers. That’s how capitalism works,” he wrote. He drew a line between commercial failure—which he believes should play out in the market—and catastrophic risks where only government-level action could be effective (for example, a nation-scale AI-driven cyberattack). In that latter domain, Altman accepts a role for the state as an insurer of last resort — but not as a financier of private buildouts.

Altman’s third answer addressed timing: why front-load enormous infrastructure spending instead of scaling more gradually? His argument is operational. Massive compute projects take years to build; demand is already outpacing supply; and under-provisioning risks hamstringing the very innovations AI promises to unlock. “The risk to OpenAI of not having enough computing power is more significant and more likely than the risk of having too much,” he wrote, pointing to current rate limits and delayed features as proof.

Altman’s post is both a commercial strategy and a policy prescription. For governments, the message is: define strategic assets, use targeted instruments (offtake agreements, low-cost capital) to build them, but avoid unconditional guarantees for private players. For enterprises and investors, the implication is to plan for a future where compute is a scarce and strategic resource and where market discipline, not political favour, determines winners.

Altman closes on an aspirational note — abundant, cheap AI that benefits society — while doubling down on market accountability: “This is the bet we are making… But we of course, could be wrong, and the market—not the government—will deal with it if we are.” It’s a high-conviction stance that reframes a national policy debate about who builds the infrastructure of the AI age and how its costs and benefits should be shared.