The public is going to own a piece of OpenAI. That is not the same as benefiting from it.

There is a deal being negotiated this week in Washington that will, on its face, give the American public an ownership stake in OpenAI. Donald Trump confirmed it from Air Force One on Friday. Sam Altman met Bernie Sanders earlier the same week. The framing on offer, from both ends of a political spectrum that agrees on almost nothing else, is that this is a win for ordinary people. I do not think that is what is actually going on here.

The convergence is real, and that is what should make you suspicious. Trump is talking about a voluntary equity stake modelled on the government's roughly 10% position in Intel. Sanders has introduced legislation requiring a mandatory 50% stake for any AI company taking federal contracts or support. OpenAI, in April, floated its own structure — a "Public Wealth Fund" that would notionally distribute returns to the American public. Three different mechanisms, three different political logics, all converging on the same surface conclusion: the public should own a piece of this.

Start with what equity actually is. A share is a claim on residual cash flows and, depending on the class, a vote. It is not a seat for the people whose work is being restructured by the thing the company sells. If the United States government acquires, say, 10% of OpenAI through a Public Wealth Fund, the dividend that flows back to taxpayers, whenever it flows, in whatever amount, is a transfer payment. It does not reach the call-centre worker before her job is restructured. It does not reach the paralegal before the firm cuts its junior associate intake. It arrives, if it arrives, after the displacement, and it arrives as a thin slice of an aggregate return rather than as compensation tied to harm.

This matters because the political case being built around the deal trades on the opposite implication. The phrase "Public Wealth Fund" is doing heavy lifting. It suggests Norway's sovereign wealth fund, which pays a citizen-facing dividend on oil revenues extracted from a finite national resource. OpenAI is not a finite national resource. Its value is being built, in significant part, on labour that is being made cheaper or redundant by its products. A dividend funded by that dynamic is not redress for the dynamic. It is a small kickback from it.

The Intel comparison is the wrong precedent, and I think Trump knows it. The government's Intel stake came out of the CHIPS Act in 2022 — a manufacturing subsidy tied to semiconductor supply-chain resilience and national security. The state put cash in, took equity out, and did not acquire governance rights over Intel's product decisions. That arrangement made sense as a tool for one job: bringing fab capacity back onshore.

An OpenAI stake is structurally different. It is being acquired at a roughly $300 billion valuation, by mechanism not yet disclosed, from a company that has just completed a for-profit conversion (the restructuring from a nonprofit-controlled capped-profit entity to a public benefit corporation) and is targeting a September IPO. There is no fab to build. There is no supply chain to onshore. The thing the public would be buying into is the upside of a product whose downside, labour displacement, market concentration, accountability gaps, falls on a different set of people from the ones who would hold the shares.

The dividend arrives, if it arrives, after the displacement.

The IPO timeline is the part of this story that is doing the most work and getting the least attention. OpenAI's September target means any unresolved political risk has to be settled before the roadshow. A live Sanders bill threatening mandatory 50% equity is a material disclosure problem. A Trump administration making noises about a stake at unclear terms is another. Altman meeting Sanders on Monday and Trump's CEOs the following week is not, primarily, a public-interest negotiation. It is pre-IPO liability management. The voluntary Public Wealth Fund proposal, floated by OpenAI in April, is the company's way of offering a controlled version of the thing both sides are demanding, before either side imposes an uncontrolled version.

That is a rational move by OpenAI. It is not, on its own, a scandal. But it should change how the deal is read. The framing being offered, "the public gets a stake", is being produced by the party with the most to lose from a worse outcome, and accepted by political actors who each get to claim a victory. The structural test of whether a deal is good for ordinary people is not whether all three parties at the table call it good. It is whether the people not at the table are better off.

And they are not at the table. The Altman-Sanders meeting and the forthcoming Trump-CEO session involve a founder whose company is on track to be one of the most valuable in the world, a senator, and a president. The workers whose jobs are being restructured by the product — customer service, junior legal, content moderation, administrative labour, increasingly mid-tier coding — are absent. The communities where the costs of deployment land most heavily are absent. There is no labour seat. There is no civic seat. There is a finance seat, two political seats, and a press release in the offing.

This is the part where I am supposed to offer a constructive alternative, and I want to be careful here, because the shape of the alternative matters. A dividend-paying fund is better than no fund. A 10% stake produces more taxpayer return than 0%. I am not arguing against a public stake. I am arguing against accepting a public stake as a substitute for the things it does not do.

The things it does not do, specifically: it does not give displaced workers a claim on the productivity gains their replacement generates. It does not give affected communities consultation rights on deployment. It does not give the government governance rights over which products ship, on what timeline, with what safeguards. It does not address market concentration. It does not change who decides. It produces a financial return and calls it democratic participation.

The honest version of this deal would acknowledge what it is: a financial settlement between a company that needs political cover before an IPO and political actors who need a win. The dishonest version is the one being staged this week, in which an ownership stake is presented as the public getting its due. The public is not getting its due. The public is getting a small share of the upside of a transition whose downside it will continue to bear in full.

I will be watching the September prospectus. The structure disclosed there will tell you which version of the deal actually closed.

Footnotes

¶