The Taxpayer Bailout of AI

Day 21 of the 1,000-Day Horizon

Either it works — or it doesn’t.

And yet, before either verdict is in, we’re already being told who will pay if the bets go bad.

The CEO of OpenAI, Sam Altman, says it plainly, almost casually:

“At some level, when something gets sufficiently huge… the federal government is kind of the insurer of last resort.”

That sentence is doing a lot of work.

It treats a taxpayer backstop as inevitable rather than optional. It quietly removes bankruptcy from the conversation. And it reframes speculative corporate risk as a public obligation — before AGI has proven it can carry its own weight.

We’ve seen this movie before.

The 2008 financial crisis — triggered by leverage, opacity, and the assumption that the system would be rescued — ultimately cost taxpayers on the order of a trillion dollars in direct and indirect support. And that was a housing and credit shock.

An AI bailout would be bigger by design: concentrated compute infrastructure, massive energy build-outs, national labor displacement, and firms deemed “too central” to fail.

Once you socialize downside at that scale, you’re not talking about billions. You’re talking ten to twenty trillion dollars or more over time — quietly absorbed through debt, inflation, and foregone public investment.

Let’s slow this down.

Outcome One: AGI Works

If AGI works the way its advocates claim — sustained productivity gains, real scientific breakthroughs, and large-scale labor displacement — then yes, the social contract has to change.

People will lose jobs. Entire categories of work will compress or disappear. In that world, guardrails are for people, not firms.

Income stabilization. Transition support. Education that actually leads somewhere. Public investment in resilience. Society absorbs the shock because the gains are real and broadly shared.

That isn’t socialism.

That’s capitalism acknowledging its blast radius.

Outcome Two: AGI Is a Bubble

But if AGI doesn’t work — if the economics don’t close, the breakthroughs stall, or the returns fail to justify the capital — then this is just another speculative cycle.

And speculative cycles have rules.

Bad bets fail.

Companies restructure or go bankrupt.

Investors eat losses.

Assets get repriced.

Taxpayers do not step in to save hype.

This is the part Altman glides past.

When Tyler Cowen presses him on the nuclear analogy — a sector where the government explicitly backstops catastrophic risk — Altman pulls back:

“I do think the government ends up as the insurer of last resort, but I think I mean that in a different way than you mean that… I don’t expect them to actually be writing the policies.”

That “different way” matters.

Because nuclear power didn’t get a blank check. It got a tightly defined framework: private insurance first, industry pooling second, and only then a capped, explicit public backstop.

What’s being normalized for AI is something looser and more dangerous: an implied guarantee.

Not written down. Not voted on. Just assumed — because the technology is “too big” to fail.

The Thing Being Erased

Capitalism without bankruptcy isn’t capitalism.

Bankruptcy is the firebreak. It’s how markets learn. It’s how excess gets burned off instead of socialized. Remove it, and you don’t get innovation — you get leverage without discipline.

Once firms believe the downside is public, speed beats safety. Scale beats caution. “Move fast” becomes a liability transfer.

That’s not inevitability.

That’s moral hazard by pre-announcement.

Altman gestures at the broader shift:

“I do put a significant probability that the social contract has to change significantly.”

Maybe it does. But changing the social contract for people is not the same as indemnifying corporations in advance.

One is democratic.

The other is corporate socialism.

Draw the Line Now

If AGI is real, it will earn public trust by delivering results — but we begin debating now how society cushions the human cost.

If it isn’t, the market must be allowed to clear.

No insurer of first resort.

No implicit guarantees.

No taxpayer bailout of speculative compute bets.

Two outcomes.

One clock.

Day 21.