Jensen Huang names Anthropic as customer

Jensen Huang used Nvidia’s latest earnings-call commentary to put a name on one of the company’s newest frontier-lab customers: Anthropic. On May 20, the Nvidia chief executive said the company had “added Anthropic to our partnership this year” and was bringing substantial new computing capacity online for the AI lab across multiple cloud providers. Huang tied that expansion to a broader claim that Nvidia is gaining share in inference, the stage of AI computing used to run deployed models rather than train them. His remarks came two days after Nvidia said it had delivered its first Vera CPU systems to Anthropic and other large AI customers. ### Where did Huang say Anthropic had become a customer? Christopher Muse of Cantor Fitzgerald asked Huang on Nvidia’s May 20 earnings call how Vera Rubin and “extreme co-engineering” with frontier models would affect Nvidia’s inference market share in late 2026 and 2027, according to a transcript carried by 24/7 Wall St. Huang answered that the number of frontier model companies had grown and said, “we added Anthropic to our partnership this year.” (247wallst.com) Huang said Nvidia had partnered with Anthropic to secure computing capacity across Azure, AWS and CoreWeave, adding that “the amount of capacity that we’re going to bring online for Anthropic this year and next year is going to be quite significant.” He also said Nvidia’s prior “coverage of Anthropic has been largely 0 until just recently.” (247wallst.com) ### What exactly is new here if Anthropic was already tied to Vera Rubin? Nvidia had already named Anthropic publicly in March when it launched the Vera Rubin platform at GTC in San Jose. In that March 16 announcement, Anthropic CEO Dario Amodei said Vera Rubin would give Anthropic “the compute, networking and system design” needed to support more complex reasoning and agentic workloads. (247wallst.com) May brought a more operational update. Nvidia said on May 18 that its first Vera CPU systems had arrived at Anthropic’s San Francisco office, alongside deliveries to OpenAI, SpaceXAI and Oracle Cloud Infrastructure. Nvidia Vice President Ian Buck hand-delivered the systems, and Anthropic head of compute James Bradbury said the company was “excited to see Vera emerge as a promising part of the ecosystem when solving for agentic workloads.” (hpcwire.com) ### Why did Huang connect Anthropic to inference share? Huang’s answer was framed around inference, not training. On the call, he said Nvidia was “growing share in inference very, very quickly” because the number of frontier model companies had increased and because Anthropic’s demand was expanding fast. (blogs.nvidia.com) Nvidia has been pitching Vera Rubin as infrastructure for both training and real-time AI serving. Its March product announcement described the platform as designed for “every phase of AI,” including pretraining, post-training, test-time scaling and “real-time agentic inference.” ### What did Huang say about Vera Rubin adoption across labs? (247wallst.com) Huang told analysts that “every single frontier model company will jump on Vera Rubin from the get-go,” adding that this had not been true for Blackwell. He said Vera Rubin would be “more successful than Grace Blackwell.” (hpcwire.com) Nvidia’s March launch materials named Anthropic, OpenAI, Meta and Mistral AI among labs planning to use Vera Rubin. The company said the platform combined the Vera CPU, Rubin GPU, NVLink 6, ConnectX-9 SuperNIC, BlueField-4 DPU, Spectrum-6 Ethernet switch and Groq 3 LPU into a rack-scale system for AI factories. ### What happens next for Nvidia and Anthropic? (247wallst.com) Nvidia has said Vera Rubin is already in full production and that the “road to Vera-powered systems” has begun, with initial deliveries made in mid-May. Huang said additional Anthropic capacity would come online this year and next year across cloud partners including Azure, AWS and CoreWeave. (blogs.nvidia.com) (hpcwire.com)