Five Percent of Nothing

OpenAI has proposed giving the US government a 5% equity stake in the company [1]. At an $852 billion valuation, that stake would be worth more than $40 billion. The talks have been going on for over a year, according to the Financial Times, and remain at a "conceptual" stage [2].

The headline number is eye-catching, but the real story is what the proposal reveals about the increasingly tangled relationship between frontier AI labs and the state.

The Alaska model

Sam Altman has reportedly pitched the idea as an AI-era sovereign wealth fund, modeled on the Alaska Permanent Fund [3]. The Alaska fund invests state oil revenue and pays dividends to residents. The analogy is straightforward: AI is the new oil, and the public should get a cut.

Under the proposal, all leading US AI developers, not just OpenAI, would contribute 5% of their shares to a shared vehicle. Altman has negotiated directly with President Trump, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Bessent [4]. He also spoke with Senator Bernie Sanders, who wants far more: a 50% stake in every US AI company through his own sovereign wealth fund legislation [5].

Putting the plan into action would likely require an act of Congress, which significantly complicates matters [6].

What is really being bought

The proposal is framed as public benefit, but you can read it several ways.

First, as genuine public stake. If AI drives the kind of economic disruption that labs themselves predict, distributing equity makes sense. OpenAI's own April policy paper, "Industrial Policy for the Intelligence Age," argued that returns from a public fund "could be distributed directly to citizens, allowing more people to participate directly in the upside of AI-driven growth" [7]. That sounds noble in theory.

Second, as political insulation. The timing is not accidental. OpenAI is under mounting political pressure, from cybersecurity concerns to fears of AI-driven unemployment. Once the government owns a stake, the political calculus shifts. A senator who might want to regulate OpenAI aggressively has to weigh the effect on the government's own investment [8].

Third, as a bailout backstop. If OpenAI's finances go south, and there are real questions about the sustainability of frontier model economics, having the government as a shareholder makes a rescue more likely. The taxpayer becomes the safety net [9].

All three readings can be true simultaneously, and that is what makes the proposal interesting. It is philanthropy, lobbying, and risk management wrapped into a single pitch.

The Sanders counter-proposal

Senator Sanders introduced the American AI Sovereign Wealth Fund Act in June, calling for a one-time 50% tax on AI company stock [10]. The bill would apply to all "systemically important" AI companies, including those dealing with data centers, infrastructure, or robotics. Companies like Google and SpaceX that include AI as only part of their business could spin off non-AI portions to avoid taxation.

The bill has yet to advance to committee, and at 50% it is likely a negotiating position rather than a realistic demand. But it establishes the left endpoint of the debate. The right endpoint is "no government stake at all." OpenAI's 5% sits much closer to the right than to Sanders, which is probably the point.

Why this matters if you are not American

If you build on AI APIs anywhere in the world, the governance of AI labs affects you. Three weeks ago I wrote about Anthropic's Fable 5 being shut down by US export controls [11]. That lasted three weeks. The models came back with government pre-release access, safety classifiers, and ongoing oversight built in.

A government equity stake is a different mechanism but the same direction: the state gets embedded deeper into the AI stack. Today it is a 5% stake negotiated voluntarily. Tomorrow it could be a 50% tax mandated by law. Either way, the frontier models the rest of the world depends on are increasingly shaped by US domestic politics.

The Alaska fund works because oil is a commodity. AI is not. A barrel of oil is the same whether the state owns 5% or 50% of the well. But a model trained under government oversight, with safety classifiers designed to satisfy federal agencies, is a different product than one developed independently. The stake does not just buy a financial return. It buys influence over what gets built and how.

The real question

The debate is being framed as "how much should the public get?" The better question is "what does the public get?" A 5% stake worth $40 billion sounds like a lot of money, but if it comes with reduced regulatory pressure and a tacit government backstop for a single company, the public might be selling something more valuable than they realize.

OpenAI declined to comment to the Financial Times. The White House did not initially respond to a request for comment [12]. The talks are preliminary. But the direction is clear. The AI industry and the government are not just regulating each other anymore. They are merging.