Anthropic CEO described Jensen Huang as 'kind of Trump-like' during 2022 meeting
Nvidia CEO Jensen Huang and Anthropic CEO Dario Amodei had a rocky first meeting, according to an upcoming book. Kevin Roose, the author of the forthcoming book, The AGI Chronicles, published an excerpt on X that their initial rendezvous was filled with disagreements and included some slights.
The meeting, which the book says took place in May 2022 at the high-end Chinese restaurant Eight Tables in San Francisco, was to discuss Anthropic's compute needs.
Apropos of Jensen/Ezra, here's a story from The AGI Chronicles about the first time Dario Amodei and Jensen met to discuss Anthropic's compute needs. Didn't go well! pic.twitter.com/KQxsHarE9ySeptember 24, 2026
Roose writes that ahead of the dinner, Anthropic's chief compute officer, Tom Brown, had attempted to haggle for a discount on GPUs, showing Huang a spreadsheet suggesting that Tensor Processing Units from Google were, "dollar-for-dollar, a better investment than Nvidia's chips." That's when things reportedly got nasty.
The book claims that Huang, angry at the comparison, called Brown a "bean counter," though an Nvidia spokesperson denied that to Roose.
The dinner afterward didn't go much better. Huang reportedly talked about building the biggest data center in the world, which Amodei repeatedly pressed Huang on with various questions. Huang "just kept repeating himself," the book alleges, citing a source who saw the dinner. That led Amodei to mutter that Huang was being "kind of Trump-like." While no deal was struck at the dinner, the companies did later come up with a partnership.
Today, Anthropic uses a mix of GPUs from Nvidia and AMD, Google's TPUs, and is working on co-designing its own custom inference chips. Amodei recently published an essay entitled "We Must Pace the Frontier" recommending slowing down the pace at which new AI models are improved and dined with the actual Donald Trump.
Huang has publicly disagreed with calls for pacing and regulation, and instead has said in interviews that companies should self-police. Huang also said that companies that release unsafe products should be shut down. Many rogue AI attacks, such as OpenAI's model attacking Hugging Face, were in training during these incidents.
The AGI Chronicles releases on Oct. 6.