Trump: US May Take Equity Stakes in OpenAI, Anthropic

The United States government may soon own a piece of OpenAI. That is not a rumor from a fringe newsletter — it is what the president told TIME magazine, on the record, in the White House. Asked why Washington shouldn't take stakes in the frontier AI labs the way it did with Intel, Donald Trump answered: "I might. Maybe I could do that."

He said no to nationalizing them. He said the Justice Department is the only guardrail they need. And then he spent the next several minutes explaining, in considerable detail, how much money the government could make as their shareholder. The interview was conducted on September 28 and published October 1. Nobody at OpenAI or Anthropic has commented since.

Photo: President Donald Trump at the Resolute Desk in the Oval Office, with stacks of executive orders and aide Stephen Miller. In his September 28 TIME interview, Trump said the US "might" take equity stakes in OpenAI and Anthropic. (Official White House photo, public domain.)

The "I might" heard around the AI world

The exchange is worth quoting at length, because the transcript — a 75-minute sit-down with TIME senior political correspondent Eric Cortellessa and editor-in-chief Sam Jacobs — reads like a man talking himself into an idea in real time.

TIME asked whether Trump would consider nationalizing the top frontier labs, given that he calls the AI race essential to national security. "No, no," he said. Pressed on why not, he pointed to his law-enforcement backstop: "And I have a guardrail. You know what the guardrail is? The Department of Justice. If they do bad things, I have a DOJ, I have an FBI. That's your guardrail over here, and I've used it twice, and it was very effective."

Then TIME pivoted to Intel. "You've invested in Intel," the interviewers noted. "You're taking a piece of Nvidia sales... why let someone else run it?" Trump launched into the story of the Intel deal as he tells it — the CEO came to see him, he had a problem Trump could solve, and Trump said, in effect: I'll fix it, but the United States of America gets 10 percent of your company. "He said, 'You have a deal.' After that was done, the deal was done, the stock went through the roof, you know."

"I think we made $60 billion," Trump said. And when TIME asked the obvious follow-up — why not do that with OpenAI and Anthropic — the president didn't hesitate. "I might. Maybe I could do that. I have many deals like that — I do that. But I don't get anything because I made that deal... I made it for the United States."

He kept going. "Actually, it's super American," he said, answering critics who call government equity stakes un-American. "But we made like $60 or $70 billion. Intel's the hottest stock there is, just about. Right at the top." And then the kicker, the line that will be quoted back at him for months: "I do that all the time. I don't do it for myself. I do it for the country."

The Intel stake math doesn't add up

We ran the share math ourselves, because the entire argument rests on it. The Intel deal, announced in August 2025: the US government received 433.3 million shares — a 9.9 percent stake, marketed as 10 percent — for $8.9 billion. The money wasn't new spending; it was $5.7 billion in previously awarded but unpaid CHIPS Act grants plus $3.2 billion from the Secure Enclave program, converted into equity. The stake is passive: no board seat, no governance rights, no information rights.

433.3 million shares at the $20.47 deal price is $8.87 billion. That part checks out. The "$60 or $70 billion" does not.

At Intel's September peaks, the stake's total value touched roughly $52–62 billion — but that is the value of the holding, not the profit. The unrealized gain, the actual money "made," was somewhere between $35 and $44 billion depending on the date you pick. Trump conflates the two, and he has done it before: in a June Truth Social post he wrote that "America's stake is now over 60 Billion Dollars," describing the stake's market value as though it were profit. Nothing has been sold. It is paper.

This matters because the pitch to take stakes in OpenAI and Anthropic is, at bottom, a pitch about returns. If the exhibit-A example is inflated by tens of billions of dollars, the salesmanship deserves the same scrutiny as the policy.

Why OpenAI and Anthropic — and why now

The idea didn't come out of nowhere. In July, the Financial Times reported that OpenAI had held early "conceptual" talks about giving Washington a 5 percent ownership stake — worth something like $42 billion at the company's valuation at the time — with CEO Sam Altman personally pitching Trump, Commerce Secretary Howard Lutnick and Treasury Secretary Scott Bessent. The proposal reportedly extended to Anthropic, Google and Meta contributing stakes into a government-run investment vehicle modeled on the Alaska Permanent Fund. Altman had first floated the concept to Trump directly in early 2025.

Anthropic, notably, was never in those talks — at least not according to a sourced Reuters denial in July, which said the administration and Anthropic had "not discussed the government taking stakes in the firm." The White House, the Commerce Department and Anthropic all declined to comment further at the time.

What changed since is the relationship. The TIME interview took place on September 28 — one day after Trump hosted Anthropic CEO Dario Amodei for a private two-and-a-half-hour White House dinner. That's a remarkable thaw for a relationship that, in February, featured Trump attacking Anthropic's "left wing nut jobs" on Truth Social, the Pentagon designating the company a supply-chain risk, and agencies being told to stop using its models.

And the day after the interview, September 29, Trump had Amodei and five other tech chiefs — Sundar Pichai, Elon Musk, Mark Zuckerberg, OpenAI president Greg Brockman, Jensen Huang — at the White House signing the voluntary AI safety pledge Trump called "morally binding." The choreography is hard to miss: dinner, interview, pledge, all within 72 hours. The administration is wrapping its arms around the frontier labs in public while musing, in the same breath, about owning pieces of them.

There's also the IPO clock. Reuters reported October 2 that Anthropic's IPO prospectus — the filing that locked in $518 billion of uncancelable compute commitments — for an offering that could value the company at $2 trillion — explicitly warns that US government attitudes toward the company could cause "material revenue losses or business disruptions," citing the February model ban and the Pentagon's supply-chain-risk designation. A government that is simultaneously your regulator, your potential shareholder, and a named risk factor in your prospectus is a new kind of relationship for American business.

These are the same frontier labs whose models power the agents — and AI agents are already building entire Shopify stores.

Anthropic's November IPO changes the whole calculation

Update, Oct. 3: Since this story published, the stakes Trump floated got a deadline. Bloomberg reports (via Asianet Newsable, AI Industry Today) that Anthropic has sent invitations for an October 14 pre-IPO investor day at its San Francisco headquarters, with formal IPO marketing as early as the week of November 9 and a target of trading on Nasdaq before Thanksgiving on November 26. Prospective investors are discussing a $1.8–2 trillion valuation — not an official price target — with Morgan Stanley, Goldman Sachs and JPMorgan leading an offering expected to raise around $100 billion. If Washington wants a piece of Anthropic now, it can't be a quiet backroom deal: it would have to buy in public, at market prices, alongside everyone else. That makes Trump's "I might" a lot more complicated than it sounded on September 28.

"The Department of Justice" is the guardrail

Buried in the stakes discussion is arguably the more consequential line. Asked what restrains the labs building systems their own makers warn could pose catastrophic risks, Trump's answer was not a regulator, not a law, not an agency — it was prosecutors. The DOJ. The FBI. Used twice, "very effective."

TIME also asked about reports that OpenAI's agents had breached federal government websites. "They're not allowed to do that," Trump said, "and if they do that, they, you know, could have penalties that are not going to be acceptable to them."

Consider what that worldview implies: the most powerful technology in human history, governed by the threat of after-the-fact prosecution. No speed limits, just traffic court. It is of a piece with everything else this administration has done on AI — the executive order rebranding AI as "Super Intelligence" ("It's not artificial intelligence because artificial intelligence is like CNN. It's fake," Trump told TIME, a line so perfectly him it almost reads as parody), the voluntary pledge with no enforcement mechanism, the vow of no slowdown. The labs can move as fast as they want; the DOJ will sort it out later.

While the White House calls the DOJ the only guardrail AI needs, Congress is reaching for handcuffs — Senators Hawley and Murphy are threatening prison time for AI agent hacks.

Could Washington actually buy in? The legal roadblocks

Then there's the question nobody in the White House has answered: can the government actually do this? The Intel deal is the closest precedent — a 9.9% stake worth about $8.9 billion, assembled from $5.7 billion in CHIPS grants and $3.2 billion from the Secure Enclave program, which Intel described as passive: no board seat, no special governance or information rights, with the government voting alongside the board. But Intel is a public company with a stock price. OpenAI is still private, and Anthropic is weeks from an IPO — neither has commented on Trump's remark, and neither has announced any deal or timetable. Until someone explains the mechanism, "I might" is a wish, not a plan.

The pushback: "corporate-government fusion"

As of this writing, neither OpenAI, nor Anthropic, nor the White House has responded on the record to the stakes remark. The silence from the two companies is deafening — and understandable. What do you say when the president muses about buying a piece of you?

The criticism that does exist comes from earlier this year, when the Intel deal was fresh, and it hasn't aged. David Sacks — Trump's own former AI czar — warned: "America won't win the AI race if we beat China but end up with a CCP-style social credit system in the U.S.," calling government equity stakes "corporate-government fusion." Senator Thom Tillis, Republican of North Carolina, told CBS: "For so many of my self-described true conservatives, you're going to have to explain to me how this reconciles with true free-market capitalism." Short-seller Jim Chanos put it more bluntly: "What do you call it when both Trump and Sanders advocate for government ownership of corporate equity?" (Senator Bernie Sanders, for his part, counter-proposed a 50 percent one-time tax on the largest AI companies' shares.)

The through line of the objections is the conflict: the government as shareholder, regulator, customer and — per the TIME interview — prosecutor, all at once. When the state owns 10 percent of Intel and takes 15 percent of Nvidia's China chip revenue (the H20 arrangement, confirmed last August), "I do it for the country" starts to sound less like industrial policy and more like a business model. As one tally has it, the US has taken stakes in at least 30 companies since last summer. The department store is open, and the owner is shopping.

Our take

Let's be clear about what happened here. The president of the United States sat for a marquee magazine interview and floated the federal government taking ownership stakes in the two most important AI companies on earth — casually, almost as an aside, with numbers that don't survive contact with a calculator. Then everyone moved on to the next question.

That casualness is the story. A decade ago, the suggestion that Washington might nationalize-by-equity-stake the country's leading technology firms would have been a five-alarm political fire. Now it arrives sandwiched between a dinner invitation and a voluntary pledge, and the markets don't even blink — partly because both companies are private, partly because we've all been desensitized by the sheer volume of it. Intel. MP Materials. U.S. Steel's golden share. Nvidia's revenue cut. Each one normalized the next.

The honest case for the policy is not crazy: if taxpayers are going to underwrite the AI buildout — through grants, export licenses, energy policy, and the implicit backstop of American power — why shouldn't taxpayers own a piece of the upside? The Alaska Permanent Fund analogy has genuine populist appeal, and "super American" is, whatever you think of the messenger, a defensible description of a sovereign wealth instinct most developed countries already indulge.

But the honest case against it is stronger, and it fits in one sentence: you cannot be the referee, the player, and the house at the same time. A government that owns OpenAI cannot credibly regulate OpenAI; a Justice Department that answers to the president cannot be the "guardrail" for companies the president owns pieces of. The conflicts don't disappear because the intentions are declared patriotic. They compound.

Whether this ever becomes more than an interview riff is unknowable — Trump announced no talks, no terms, no timeline. But the Overton window just moved. The next time a "conceptual" 5 percent offer circulates, it won't sound conceptual at all. It'll sound like the opening bid.

Europe isn't waiting to see how this plays out — Bull is doubling output at Europe's only supercomputer factory as part of its own sovereignty push.

Sources: TIME interview transcript (Sept 28, published Oct 1; via full-text mirror); Reuters (Anthropic IPO prospectus, Oct 2; July stake-talks denial); Financial Times (July OpenAI 5% report); CNN; People; Coinpaper; TokenPost; NY Post; Bloomberg.

By Daniel Rao, TechPulse USA

Daniel Rao covers AI, labor economics and US technology policy for TechPulse USA.