U.S. Government Negotiates Equity Stake in OpenAI: Trump Confirms Talks About 'Public Wealth Fund'

Confirmed on Board Air Force One

The Trump administration and OpenAI are in negotiations for the U.S. federal government to acquire an equity stake in the company, valued at over $850 billion by private investors. President Donald Trump confirmed the discussions on June 5 while speaking with reporters on Air Force One: "There are concepts in which shares could be given to the American people so that the American public would essentially become a partner." Axios published the full context of the presidential statement on June 6; CNBC had reported the existence of the negotiations a day earlier.

The framework under discussion was described by OpenAI itself in a policy proposal published in April 2026. The company would transfer shares to the government, which would deposit them into a "Public Wealth Fund" tasked with investing in "diversified long-term assets" and eventually distributing returns to American citizens. No official terms had been agreed upon by the time of this report, and OpenAI had not publicly commented on the negotiations.

Two Bills, Two Opposing Models

The idea that the U.S. government should hold a stake in the largest AI companies gained, in June 2026, two advocates with radically distinct architectures. The model being discussed with the White House is voluntary: OpenAI decides the volume of equity to transfer, and the government does not acquire management rights. There is precedent: during his second term, the Trump administration took stakes in companies such as Intel and IBM through strategic technology contracts.

Senator Bernie Sanders introduced the American AI Sovereign Wealth Fund Act to Congress, proposing that companies like OpenAI, Anthropic, and xAI be required to give up 50% of their shares to the government through an equity tax. The resulting sovereign fund would grant the government voting rights and equal seats on the boards of directors; the returns would be distributed as a universal dividend and, in the long term, would fund public healthcare, education, and housing.

In Trump's proposal, OpenAI remains a private company that chooses to become a partner with the state. In Sanders' proposal, it loses half of its equity control by decree. The practical difference is significant for the IPO timeline of companies in the sector: Anthropic confidentially filed an S-1 with the SEC on June 1, after closing a $65 billion round at a valuation of $965 billion. Any company in this position tends to prefer the voluntary version.

The Timing and the IPO Race

Sam Altman discussed the idea of government involvement directly with Trump as early as 2025 and revisited the subject with senior administration officials throughout the second term. OpenAI, valued at over $850 billion by private investors, plans to go public in 2026. Any federal stake established before the IPO would have direct implications for the shareholding structure and corporate governance mechanisms during and after the public offering.

The negotiations also involve Anthropic, according to Cryptobriefing. The company had not commented on any potential discussions with the government by the time of this report.

Three Models, Two Continents

The American proposal does not operate in a vacuum. The European Union opted for mandatory regulation, not equity participation: the AI Act comes into full effect on August 2, 2026, with transparency requirements, risk assessments, and restrictions on high-risk AI systems. The European Commission obtained access to Anthropic's Claude Mythos not through equity, but through a diplomatic agreement confirmed by CNBC on June 1, following weeks of government-mediated negotiations with the U.S.

In Japan, a third pathway: state-to-state diplomacy, without capital and regulation. Japanese Finance Minister Satsuki Katayama confirmed that MUFG Bank, Sumitomo Mitsui Banking, and Mizuho Bank received access to Claude Mythos and OpenAI’s GPT-5.5-Cyber. This access was facilitated by U.S. Treasury Secretary Scott Bessent's visit to Tokyo in May 2026. The Japanese government does not intend to become shareholders in these companies; it negotiates for preferential access to models that the open market cannot acquire yet.

The three models reveal that the debate about AI governance often raises the wrong question. The dispute is not just about who controls the algorithms, but also about who controls access to them when the most powerful models are treated as strategic state assets. For a CIO who needs to justify to the board which AI infrastructure the company may engage in over the next 24 months, Washington's trajectory in relation to OpenAI is as relevant as the product roadmap of suppliers.