Sam Altman has proposed that the U.S. government receive a five-percent equity stake in OpenAI — and, per the structure he has outlined to the Trump administration, in Anthropic, Google, and Meta as well — through a sovereign wealth vehicle modeled on the Alaska Permanent Fund [1]. At OpenAI's March 2026 private-market valuation of $852 billion, five percent is worth $42.6 billion [2]. The Alaska framing is the distraction. The financial consequence is the story.

The Alaska Permanent Fund works because oil revenue flows from a finite resource into a state trust that pays dividends to Alaskans. The fund has no regulatory relationship with the companies whose production it taxes: it cannot issue permits, approve models, or restrict exports. The sovereign wealth vehicle Altman is proposing would hold equity in companies that the same executive branch simultaneously licenses to deploy products, reviews for national security, and subjects to export controls. That is not a passive dividend fund. That is a financial position inside a regulatory process.

This paper's July 2 story establishing that Commerce is already an operational gatekeeper for Anthropic's model deployments described a government that acts as a standing regulator of frontier AI. Altman's proposal adds a financial dimension to that regulatory relationship: the same administration that decides whether to lift or extend an export-control order on Anthropic's models would, under the proposal, hold a 5% stake in Anthropic's valuation. A government that holds equity in a company has a financial interest in that company's success — which creates a financial interest in its competitors' failure.

The conversations have reached Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick, both of whom govern regulatory instruments that directly affect AI company valuations [1]. Altman has also raised the concept directly with President Trump [2]. The simultaneity matters: the White House is currently negotiating both the 30-day frontier-model preview framework (which determines which models require federal review before launch) and the equity-stake structure (which would give the government a financial position in the companies being reviewed). The same set of companies has leverage in both conversations at once.

Altman's proposal envisions Google, Anthropic, and Meta making parallel stake offers through the same vehicle [2]. Whether any of those companies has agreed is not public. The calculation each of them faces is identical: a five-percent equity transfer buys regulatory goodwill from an administration that already holds substantial informal power over their operating conditions, at the cost of formalizing a financial relationship with an entity that may not be a neutral regulator once it has equity interests in the outcome.

The scale of the asset changes how every future AI regulatory decision gets made. A five-percent stake at current frontier-AI valuations would make the U.S. sovereign wealth vehicle one of the largest institutional investors in the technology sector within months of its creation. The White House would not need to issue regulatory guidance favoring incumbent frontier labs over challengers — it would have a financial incentive to do so that is documented in a portfolio statement.

Senator Bernie Sanders proposed a fifty-percent tax on AI company stock in June, with the collected shares deposited into a public fund under the American AI Sovereign Wealth Fund Act [3]. The Sanders proposal treats AI equity as a public resource to be taxed. The Altman proposal treats it as a gift to secure political accommodation. Both proposals assume that a government equity position in AI companies is achievable in principle; they disagree sharply about whether that position should be earned or purchased.

Any formal structure for the stake would likely require congressional approval, which means the talks remain preliminary and the timeline uncertain [1]. But the talks themselves — taking place the week before the White House is expected to announce the frontier-model review framework — are already an operating variable. Companies that have offered stakes, or signaled willingness to offer them, enter the framework negotiation in a different position than companies that have not.

-- THEO KAPLAN, San Francisco