Nvidia released more details and specifications for its Vera data center CPU on July 21 and said the first chips had been delivered to OpenAI, Anthropic, and SpaceX.

The move expands Nvidia beyond AI GPUs into server CPUs, where it will compete with Intel and Advanced Micro Devices.

Nvidia said Vera is its first server CPU designed in-house, using 88 custom Olympus cores. The company said the cores delivered up to 50% higher instructions per cycle than Grace, its earlier data center CPU.

The company said it will sell Vera on its own and with Nvidia GPUs in systems such as Vera Rubin.

Nvidia’s more specific performance claim is narrower than a broad comparison with x86 chips.

It said Vera delivered more than 1.8x higher sandbox performance on agentic workloads than a “latest x86 CPU,” based on Nvidia-measured data, while the 50% figure referred to per-core gains vs. Grace.

Nvidia’s materials also suggested the early shipments were for evaluation rather than production. SpaceX, for example, is evaluating Vera for reinforcement learning workloads.

The move comes as AMD holds 46.2% of x86 server CPU revenue, according to Mercury Research data.

🔗 Source: CNBC