OpenAI's 5% Government Stake Isn't What It Sounds Like

Five percent of OpenAI is worth about $42.6 billion right now. That's the number attached to the stake Sam Altman reportedly floated giving the US government this week, and it's worth sitting with for a second before getting into what it actually means, because "government stake in OpenAI" sounds like something it probably isn't.

Here's what's on the table, per Financial Times reporting picked up by Forbes on July 2: OpenAI has opened talks about handing Washington a 5% equity position, and the pitch extends beyond OpenAI itself. Altman's proposal envisions Anthropic, Google, and Meta contributing similar stakes into a shared vehicle. Think of it less as a government seat on OpenAI's board and more like a sovereign wealth fund that happens to hold AI equity instead of oil revenue.

That comparison isn't mine. It's Altman's. He's been pointing to Alaska's Permanent Fund since April, when OpenAI published an industrial-policy document proposing something structurally similar: a fund that holds stakes in AI companies and distributes the proceeds to the public, the same way Alaska has cut residents an annual oil-revenue check since 1976. The fund is worth roughly $91.2 billion today. If frontier AI labs are about to be worth trillions combined, the math on "AI's Alaska Fund" gets interesting fast.

This isn't a new idea surfacing out of nowhere, either. Speculation about Washington ending up on OpenAI's cap table was already circulating back in early June, before there was a concrete number attached to it.

Why an equity stake and not a tax

The obvious question is why any of this needs a special vehicle instead of, say, higher corporate taxes on AI companies. The answer is mostly political mechanics. A tax increase requires Congress and reliably becomes a fight. An equity stake OpenAI volunteers to hand over is a negotiation between a handful of executives and an administration, and it comes with a story attached: "we're sharing the upside with the public" reads a lot better in a press cycle than "we're being taxed more." Whether it functions the same way for the average person's bank account is a separate question nobody's really answering yet.

There's also a direct precedent sitting right next to this one. The US government already took a 10% stake in Intel earlier this year, and it's already taking a cut of Nvidia and AMD's China chip sales on top of that. That deal normalized the idea that Washington can hold direct equity in a strategically important tech company without nationalizing it. OpenAI's pitch reads like an attempt to get ahead of a similar arrangement being imposed on it, rather than negotiated by it.

And OpenAI has reason to want a good relationship with this administration specifically. Washington delayed GPT-5.6's release earlier this year over national-security review concerns before eventually clearing it, and separately, state attorneys general have opened their own investigation into how ChatGPT operates. A 5% stake conversation lands differently coming from a company that just got two separate reminders of how much leverage the government has over it.

The 50% counter-offer nobody expected

The stranger wrinkle here is that Bernie Sanders is reportedly in these conversations too, and he's not interested in 5%. Sanders has pushed for a government stake closer to 50%, framing it as public compensation for the public research, energy subsidies, and infrastructure buildout that got the AI industry this far. That's not a rounding difference from Altman's number. It's a different proposal entirely, one where the government isn't a passive fund beneficiary but a majority economic stakeholder in the company.

Having both the 5% and 50% versions of this idea circulating at the same time tells you these talks are genuinely preliminary, not close to a term sheet. Multiple people quoted in the coverage describe the discussions as conceptual. Altman has been having some version of this conversation with the Trump administration since early 2025, holding direct talks with the President, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Bessent, and it hasn't produced anything concrete in over a year.

What happens if Anthropic, Google, and Meta say no

The entire framing assumes the other major labs sign on to a shared structure. There's no public confirmation that Anthropic, Google, or Meta have agreed to anything, and every incentive here points toward OpenAI benefiting the most from being first. A voluntary 5% handover looks generous only in isolation. If it becomes the baseline expectation for every AI company operating at scale, and only OpenAI actually offered it, Altman comes out of this looking like the company that got ahead of regulation everyone else eventually had to match anyway.

Anthropic is already navigating its own government-adjacent tangle this year, from Commerce Department export restrictions on its Mythos model to scrutiny elsewhere in the industry, and has reportedly floated its own, separate "digital dividend" concept rather than signing onto Altman's shared vehicle. A common public fund would, in theory, put all of that under one roof instead of four separate negotiations. Whether that's appealing to a competitor with its own competing proposal depends entirely on the terms nobody's published yet.

None of this is close to law, regulation, or even a signed agreement. It's a handful of executives and officials talking about a structure that doesn't exist yet, attached to a valuation number that will have moved by the time anything gets finalized. The number worth watching isn't 5%. It's whether a second AI lab agrees to anything at all.