NVIDIA Ships Vera CPU to Top AI Labs and OCI

NVIDIA's first custom CPU has moved past the announcement phase. The NVIDIA Vera CPU landed at Anthropic, OpenAI, SpaceXAI, and Oracle Cloud Infrastructure last week, transitioning from prototype to active customer evaluation just months after Jensen Huang unveiled it at GTC San Jose in March.

Four Deliveries, One Week Across Silicon Valley

NVIDIA Vice President of Hyperscale and High-Performance Computing Ian Buck personally carried the first Vera systems to each recipient. Anthropic, OpenAI, and SpaceXAI received theirs on Friday; OCI followed on Monday at its Santa Clara AI Customer Excellence Center.

At Anthropic's San Francisco offices, James Bradbury, the company's head of compute, took the handoff. "Scaling compute is an important accelerant for the growth of models," Bradbury said. "We're excited to see Vera emerge as a promising part of the ecosystem when solving for agentic workloads."

At OpenAI's Mission Bay headquarters, Sachin Katti, who oversees compute infrastructure, received the system. SpaceXAI's stop in Palo Alto included a walkthrough with Elon Musk, who pressed Buck on core counts, memory layout, and thermal design. The company is evaluating Vera for reinforcement learning and agent-based simulation pipelines.

Why Vera Exists and What It Handles

The NVIDIA Vera CPU was not designed to replace conventional server processors. It targets the specific demands of agentic AI: orchestration, tool-calling, long-context retrieval, agent sandboxing, and the constant stream of CPU-bound tasks that run alongside GPU inference.

The chip features 88 custom Olympus cores, 1.2 TB/s of memory bandwidth, and 50% faster per-core performance under sustained load, according to NVIDIA. Those specifications matter when AI agents run dozens of concurrent operations and response latency compounds across tool calls.

Buck explained the demand shift at OCI: "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. That's why we are seeing the demand for CPUs skyrocket."

OCI Commits to Hyperscale Deployment in 2026

Oracle Cloud Infrastructure made the largest stated commitment of any recipient. Karan Batta, who leads OCI product management, confirmed plans to deploy hundreds of thousands of Vera CPUs beginning this year, making OCI the first cloud provider to adopt the chip at hyperscale.

"Vera's architecture is purpose-built for high-throughput reasoning workloads, delivering the efficiency, density and footprint OCI needs to power the next generation of enterprise AI," Batta said.

For enterprise teams already on OCI, that deployment signals production-grade agentic AI infrastructure at a scale no competing cloud currently matches. Developers running agent pipelines on OCI could interact with Vera-backed capacity without needing to configure or select it directly.

[Analysis] The range of recipients, spanning frontier AI labs, a defense-adjacent AI company, and a major hyperscaler in a single week, suggests NVIDIA is positioning Vera as the default CPU layer of the agentic AI stack rather than a niche accelerator add-on. That framing matters for enterprises planning infrastructure investments: the CPU tier, long treated as commodity, is becoming a differentiated product in AI factory design.

Source: NVIDIA Blog