Trump says he'll meet AI chiefs on a government stake
Washington moved on the AI industry's cap table, its power bill and its chain of command in a single day — three levers, one direction.
President Trump said he expects to meet the leaders of major US AI companies at the White House, likely next week, to discuss the government taking a financial stake in their businesses. Speaking to reporters aboard Air Force One, Trump framed the idea as profit-sharing rather than nationalization: the aim is to "create almost a partnership with the American public," and "where the American people can benefit from the success of AI, the American people will like it better." He said the talks are live — "We're talking about it" — and insisted he has been weighing government investment in AI firms for a year, while not dismissing the Senate's most aggressive version of the same idea, Bernie Sanders's plan for a 50% government stake. "Where economics are concerned, we have things that aren't that far apart," Trump said. Microsoft declined to comment, and the other companies reported as attendees did not respond, so treat this as a stated intention with a meeting nobody has publicly scheduled.
The architecture is already built. The administration took a 10% stake in Intel and has taken positions in quantum and critical-minerals companies; Trump signed an executive order in February calling for a federal sovereign wealth fund, and OpenAI — which has been discussing a government stake for more than a year — has floated donating equity to seed a fund-like vehicle. The politics are just as legible: public hostility to data centers and to AI-driven job losses is rising, and an equity stake converts a critic into a shareholder. Watch the structure, not the announcement. The question that decides everything downstream is whether the public gets cash-like exposure with no governance role, or a seat that turns every model release into a political decision. David Sacks, the administration's own AI and crypto chief, has warned that government ownership invites political control and censorship — an objection raised inside the tent, not by the labs.
SemiAnalysis mapped more than 300 local and state data center restrictions and put a number on them: 2.3 GW of planned US capacity faces genuine delay — about 6% of the 38 GW the firm forecasts for 2027 delivery. The gap between the moratorium count and the megawatt count is the whole finding. Local moratoriums expose roughly 20 GW of planned capacity but delay only 1,525 MW; three projects account for essentially all of it. New York's executive order touches about 1.4 GW and meaningfully delays around 0.8 GW, while Texas adds three to four months of administrative drag to grid-connected projects and simultaneously speeds up behind-the-meter builds.
Four mechanics explain the shrinkage: county moratoriums generally apply only to unincorporated territory, so a campus inside an incorporated city sits outside the ordinance; they freeze new applications rather than revoke existing approvals, which leaves all of 2026 and most of 2027 delivery untouched; most run three to six months and expire years before 2028-plus projects need the affected permit; and litigation, equipment lead times or financing often already gate the schedule, so the moratorium is not the binding constraint. "Counting moratoriums is not counting megawatts," said SemiAnalysis founder Dylan Patel. Thirteen state-level data center restriction bills died in 2026, almost none reaching a floor vote.
That matters because "moratoriums are killing the US buildout" has become a load-bearing claim in the AI capex debate — and this is the first attempt to price it project by project rather than counting ordinances. The political pressure behind it is real even where the capacity impact is not: a SemiAnalysis poll in August found 46% of voters hold an unfavorable view of data centers, including 24% "very unfavorable," and 46% would oppose one in their own town, against 46% who view AI broadly favorably. Supply, meanwhile, is being squeezed from the other end — we covered the forecast that AI data centers will out-burn Germany and Japan on gas by 2035.
Senior Trump administration officials met Anthropic's chief compute officer, Tom Brown, on Tuesday to discuss AI safety risks, with Commerce Secretary Howard Lutnick and Pentagon chief technology officer Emil Michael joining. Brown, who has become the company's face in Washington, spoke virtually; Anthropic declined to comment. A Defense Department official said the department "maintains contact with frontier AI companies" but that "the department's stance on Anthropic has not changed" — the blacklist and Anthropic's lawsuit over it remain live, and Lutnick has separately said the company had "gotten religion" and patched up its relationship with the government.
The meeting is the first tangible product of a two-week collision. Dario Amodei's 3,800-word essay on September 12 called for government regulation and paced development; Trump answered on September 14 that his presidency is the only safety guardrail AI needs and that slowing down serves China. OpenAI's global policy chief, Chris Lehane, then confirmed on Tuesday that his company has been engaging Anthropic and Google DeepMind for weeks: "It's better to try to work together to prioritise safety." Read tightly, the administration is negotiating safety with the lab it blacklisted while Treasury refuses the industry a liability shield — Bessent denies the labs a liability waiver and pushes open models — which suggests what the labs get is access, not relief.
What to watch: whether the White House meeting is confirmed, and whether any equity arrangement comes with a governance seat or only a dividend.
If the public owns a slice of the labs, does that make the public a shareholder or a hostage? Tell us in the comments.
Sources: BBC — Trump to meet AI leaders over US investment in their companies · Sina Finance — 特朗普下周拟在白宫召集AI行业高管 · SemiAnalysis — Everyone Says Datacenter Moratoriums Are Killing the US Buildout · LavX News — US Datacenter Moratoriums Delay Just 2.3 GW · Bloomberg via The Straits Times — Trump advisers meet Anthropic executive over AI safety · Mint — Anthropic is back on Trump's 'right side' after Pentagon feud, Lutnick says · Techmeme — Analysis: only ~2.3 GW delayed
