Donald Trump phoned Nvidia chief executive Jensen Huang in the middle of a live interview at the All-In Summit in Los Angeles on Monday morning and, on speakerphone, called the fear of AI takeover a hoax. The call put the head of the company that sells the hardware behind most advanced AI systems on the same side of the argument as the US president, two days after Anthropic's Dario Amodei asked the industry to slow the pace of model improvement.

Trump calls AI fear a hoax in live call to Nvidia's Jensen Huang

What happened on stage in Los Angeles

Huang was midway through the interview when his phone rang. He put the president on speakerphone, and Trump opened by noting that Huang can design the most complex computer chip in the world but could not figure out how to put him on speaker. The summit is hosted by the All-In podcast and its investor presenters, among them David Sacks, Chamath Palihapitiya, Jason Calacanis and David Friedberg. The room applauded the exchange, which was reported by Axios.

Trump's remarks went further than a comment on model risk. He said the robots will not be taking over and the AI will not be taking over the rest of the world, and described the whole thing as a hoax. Opposition to AI, he added, plays into the hands of people who do not want America to succeed, and those people could be political opponents or China. The call landed two days after Amodei said the industry must slow the rate at which model capabilities improve; Sam Altman matched that position within hours, and Elon Musk agreed as well.

Huang did not fully adopt the president's wording. At the same event he called apocalyptic predictions not grounded in science, a narrower claim than a hoax, and he praised Jacob Coxon, the researcher who quit Anthropic on 9 September saying the labs were gambling with our lives, crediting him with great courage. Huang also said companies should pause or pace their work if they feel their company is out of control. That position sits close to what Amodei asked for: the distinction Huang draws is between a company slowing itself down, which he can accept, and an industry-wide brake, which he cannot.

What this means for companies buying AI

For businesses, the practical consequence is that the public argument about AI risk is now split along a line that matters for procurement. A vendor that accepts voluntary self-pacing can still ship models on its own schedule, while a vendor that accepts an industry-wide pause would have to coordinate releases with competitors. When a supplier says it is slowing down, the buyer should ask which of the two it means, because only the second changes delivery dates for products already on the roadmap. For a small company, the difference shows up as a delayed API or a missing feature; for a large one, as a multi-year infrastructure plan that assumes a steady supply of new chips and models.

Several things remain unverified. Trump's claim that opposition to AI serves China has a documented kernel and a large gap around it: OpenAI reported in June that it had found a China-linked campaign using ChatGPT to stir opposition to American data centres, but its own investigator stressed the operation was interfering in an existing debate rather than manufacturing one, and an independent researcher at Clemson said his team had found little evidence of coordinated Chinese effort. Local opposition to data centres is overwhelmingly domestic, and Trump has been fighting it in Texas, New York and elsewhere for months, having told Texans in August that data centres could be bigger than oil. The news does not mean that AI development is slowing, nor that safety concerns have been settled; it means the political framing around them has changed.

One marker will show whether the shift is real: whether Nvidia's own guidance and order book hold up through the next two quarters. Nvidia sells the chips that any slowdown would leave unsold, which gives Huang the clearest commercial interest in the room of anyone discussing pace, and the president who called him was not there to test it. If data-centre orders keep growing while the talk of pacing continues, the debate stays rhetorical; if they flatten, the argument has reached budgets.