NVIDIA hand-delivers the first Vera CPU systems to Anthropic, OpenAI, SpaceXAI, and more

NVIDIA has begun hand-delivering its first custom CPU designed for Agentic AI. The Vera CPU featured 88 custom Olympus cores, an impressive 1.2 TB/s of memory bandwidth, and 50% faster per-core performance under full load. And with that, NVIDIA Vice President of Hyperscale and High-Performance Computing, Ian Buck, personally hand-delivered the first Vera CPUs to the biggest AI companies in California: Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure.

3

Although NVIDIA is synonymous with generative AI thanks to its GPUs, Agentic AI still requires a powerful CPU to handle tasks such as long-context retrieval. "Vera is a new class of CPU designed with that reality as its starting point," NVIDIA explains. "This gauntlet of concurrent, real-time tasks puts pressure on CPUs in ways traditional core-density focused designs were never built to prioritize."

As the successor to Grace, Vera underscores the growing importance of CPUs in AI, and the Vera CPU is the first data center CPU to use LPDDR5X memory, delivering 1.2 TB/s of memory bandwidth. And with that, even though it will ship as part of the Vera Rubin platform that combines CPU and GPU technology, it's also set to be made available as a standalone CPU solution. NVIDIA's hand-delivery of the first Vera CPUs signals that Vera and Rubin are entering their respective full production phases for a global rollout in the coming months.

3

"Agentic AI is creating a new CPU moment in the AI factory - as models move from answering to acting, Vera is purpose-built to keep that work moving at scale," NVIDIA's Ian Buck said. "When AI models are posed a question, the answer, often, isn't already prepped and ready to go. The models actually have to generate some Python code to arrive at the correct answer. A task at which the Vera CPU excels. That's why we are seeing the demand for CPUs skyrocket."