OpenAI Offers Washington a 5% Stake in the AI Race

Sam Altman’s equity proposal to the Trump administration reveals how deeply politics has fused with capital in Silicon Valley’s race to control frontier intelligence.

Key Takeaways

• OpenAI has proposed giving the US government a 5% equity stake worth roughly $42.6 billion, urging rival AI firms to make matching offers to a national wealth vehicle.
• The plan follows Anthropic’s brief export-control suspension and OpenAI’s own delayed GPT-5.6 launch, signs that Washington now shapes AI product timelines directly.
• Critics warn the arrangement makes government both shareholder and regulator, while Senator Sanders pushes a far more aggressive 50% equity tax alternative.

OpenAI’s Bid for Political Cover

OpenAI has proposed transferring a 5% equity stake to the United States government, a move first reported by the Financial Times and confirmed by Reuters, CNBC, and other outlets on July 2. At OpenAI’s March valuation of $852 billion, the stake would be worth approximately $42.6 billion. The company has framed the offer not as a conventional government shareholding but as a contribution to a proposed “Public Wealth Fund,” a vehicle designed to hold equity across leading AI developers and channel returns directly to American citizens.

The idea is not new. Sam Altman first raised it with President Trump in early 2025, and OpenAI formalized the concept in an April 2026 policy paper calling for shared national ownership of AI’s economic gains. What has changed is the urgency. Altman has taken the proposal directly to Commerce Secretary Howard Lutnick and Treasury Secretary Scott Bessent, while also briefing Senator Bernie Sanders, whose own legislative response sits at the opposite end of the policy spectrum.

A Sector Under Pressure

This overture cannot be separated from the regulatory climate now enveloping American AI companies. Anthropic spent several weeks last month barred from providing its most advanced models, Claude Fable 5 and Mythos 5, to customers outside the United States, after the government invoked export-control authority over national security concerns. Access was restored only this week, following changes the company made to satisfy policymakers on safety grounds. OpenAI, meanwhile, delayed the full public rollout of GPT-5.6 at the government’s request just days before the stake proposal became public.

Taken together, these episodes mark a shift in how Washington relates to frontier AI developers. The government is no longer a passive observer setting rules from a distance; it is actively pausing product launches, gating international access, and now being courted as a prospective equity holder. For an industry that spent the better part of a decade insisting it could self-govern, the change is notable. OpenAI’s proposal can be read as an attempt to convert a defensive posture into something resembling partnership, trading a slice of the company’s upside for smoother passage through an increasingly interventionist regulatory environment.

The Mechanics and the Math

The structure matters as much as the headline number. Under OpenAI’s framing, the transfer would not be a direct government stake in the traditional sense but a donation of equity into a fund that invests broadly and distributes proceeds to citizens. The company has also proposed that other major US AI developers, implicitly including Anthropic, Google, and Meta, contribute matching stakes of their own. None of those companies has responded publicly, and it remains unclear whether any intend to.

The financial backdrop makes the proposal more striking. OpenAI’s spending has reportedly reached roughly $34 billion against revenue near $13 billion, with a Microsoft compute bill alone approaching $17 billion. A company burning capital at that pace is not offering equity from a position of comfortable surplus. The stake proposal arrives alongside a confidential S-1 filing that could produce one of the largest IPOs in history, at a valuation potentially exceeding $1 trillion. Handing away 5% of the company ahead of a public listing is not a trivial concession, even framed as a gift toward the public good.

Sanders, and the Conflict Beneath the Offer

The proposal exists on a spectrum, and OpenAI’s version sits at the moderate end of it. Senator Sanders has introduced legislation, the American AI Sovereign Wealth Fund Act, that would impose a one-time 50% stock tax on OpenAI, Anthropic, and xAI, redirecting the proceeds to a government-controlled fund. Sanders has been dismissive of OpenAI’s counteroffer, characterizing a 5% contribution as a fraction of what public ownership should look like given the scale of value these companies are expected to generate. That gap, between a voluntary 5% gesture and a mandated 50% transfer, defines the real negotiation underway.

The arrangement also invites scrutiny that goes beyond politics. Public Knowledge’s Nat Purser has warned that the structure would place the government in the position of both shareholder and regulator of the same companies, a conflict with few clean precedents in American industrial policy. The Cato Institute’s Jennifer Huddleston has raised a related concern, that government equity stakes amount to picking winners in a sector still defining its competitive landscape. A government whose returns depend on OpenAI’s valuation has incentives that diverge from those of an independent regulator, and decisions about antitrust enforcement, export licensing, or safety mandates become harder to disentangle from the state’s own financial interest. The administration has taken equity positions before, in Intel, IBM, and several critical minerals and quantum computing firms, but AI’s centrality to national security and economic competitiveness raises the stakes considerably.

What Comes Next

No agreement has been finalized, and the White House, along with OpenAI, Anthropic, Google, and Meta, has so far declined to comment publicly on the details. The proposal remains, in the language of those close to the talks, an early-stage framework rather than a settled deal. Trump himself has acknowledged that discussions are underway without confirming any terms, leaving open how much of OpenAI’s framing the administration is actually prepared to accept.

Still, the fact that such a framework exists at all says something about where the industry finds itself. AI companies that once resisted government involvement are now actively designing the terms of that involvement, calculating that shaping the relationship from the inside carries less risk than having it imposed from outside. Whether OpenAI’s 5% becomes the template other frontier labs adopt, or simply a marker in a negotiation that ends somewhere closer to Sanders’ terms, will say a great deal about how the American state chooses to participate in the industry it increasingly cannot afford to ignore.