Washington is no longer only talking about regulating AI companies. It is talking about owning pieces of them, and founders should treat that as a cap table problem, not a political sideshow.
If you're building an AI company right now, the federal government may soon become an investor you never pitched. It won't look like a normal venture round. It won't behave like one either. OpenAI has held preliminary talks about giving Washington a 5% stake, according to Financial Times reporting summarized by Reuters and Axios, and the idea would also ask other U.S. AI companies to hand over similar stakes. At OpenAI's $852 billion March valuation, which the company announced after closing $122 billion in committed capital, that slice would be worth roughly $42.6 billion. No deal is signed. The point is simpler. This is now a live policy idea, not a dorm-room thought experiment. Trump pushed the door open on June 5. According to Axios, he told reporters aboard Air Force One that giving the American public pieces of AI companies could be "a beautiful thing." Sam Altman has also pushed the public-wealth-fund idea with the administration, Axios reported. You can see the appeal. AI companies get political cover while they build systems that Washington increasingly treats as national infrastructure. The public gets a claim on the upside. Everyone gets to say the future is being shared.
Look closer at the instrument. Equity is never just a slogan.
The Intel precedent is real #
The easiest version for founders to live with is the Intel model. In August 2025, Intel said the Trump administration would make an $8.9 billion investment in the company using $5.7 billion in remaining CHIPS Act grants and $3.2 billion from the Secure Enclave program. The government agreed to buy 433.3 million primary shares at $20.47 each, equal to a 9.9% stake. Intel also said the position would be passive, with no board seat, no governance rights and no information rights.
That matters on a cap table. A passive 9.9% holder is not the same thing as a director sitting in the room while you discuss model releases, licensing, safety policy or a defense contract. The government also agreed to vote with Intel's board on most shareholder matters. For a private AI company trying to raise a Series C or prepare an IPO, that version is awkward but survivable. Investors can price it. Lawyers can disclose it. Bankers can explain it.
There is still a string. Intel said Washington received a five-year warrant to buy another 5% at $20 a share if Intel stops owning at least 51% of its foundry business. That is not a board vote, but it ties a future corporate decision to a federal financial trigger. Founders should pay attention to that part, because the quiet terms are often the ones that last.
Sanders wants control, not symbolism #
Bernie Sanders' version is a different animal. His American AI Sovereign Wealth Fund Act, introduced on June 18, would impose a one-time stock tax equal to 50% of the shares of the largest AI companies. Sanders' office says the fund would be worth an estimated $7 trillion at current valuations and would be managed by an Independent Commission for Democratic AI. The commission would hold voting shares and could use them to block corporate decisions it considers harmful to the public.
That's not a passive investor. That's power.
You don't need to share Sanders' politics to understand why the proposal has oxygen. The largest AI companies are asking the public to tolerate real disruption while the value accrues first to private shareholders. If you think the models will reshape jobs, education and media, a public claim on the wealth doesn't sound absurd to many voters. A June Verasight survey reported by TechRadar found that 69% of U.S. adults supported forcing major AI companies to transfer half their stock into a public wealth fund.
That number should stop every founder cold. Reading the room through Sand Hill Road alone won't cut it any more.
Mike Bloomberg gave the cleanest objection. In a Bloomberg Opinion column republished by The Star and covered by Fortune, the former New York City mayor argued that government ownership would turn Washington from regulator into profit-seeking shareholder. He wrote that the market would become "a smoke-filled backroom" where success depends on political connections, lobbying budgets and elections. He also warned that government-owned AI raises speech risks, because the state would have both political and financial reasons to care what the systems say.
Frankly, that is the hard part of this story. Government grants have built important technology before. DARPA helped fund foundational work for the internet without asking to sit on every cap table that benefited from it. Grants can set conditions. Procurement can create markets. Equity does something else: it makes the referee part-owner of the team. That is a different relationship entirely.
None of these proposals has cleared Congress. OpenAI and the White House did not immediately comment to Reuters on the Financial Times report - and that dry detail belongs here, because this is still a proposal stack, not a signed term sheet. But the Intel deal proves the administration can convert public money into shares of a major technology company without waiting for a clean new statute built for that purpose.
For founders, the lesson is not to panic. It is to model the terms. A 5% public-wealth stake, a passive 9.9% federal stake and a 50% voting transfer are not variations of the same idea. They produce different boards, different exits and different conversations with every future investor. If Washington wants AI upside, you need to know whether it wants a dividend, a veto or a seat just outside the boardroom. Also read: Meta's AI is banning thousands of innocent Facebook and Instagram accounts and offering no real way back • Big Tech is spending $725 billion on AI in 2026 and its free cash flow is nearly gone • TSMC's record AI chip profits come with a hidden tax that every founder should understand