SpaceX Receives First NVIDIA Vera CPUs; Musk Praises Chip

NVIDIA confirmed it delivered early Vera silicon to SpaceX, OpenAI, Anthropic and Oracle Cloud Infrastructure as part of an initial rollout. The company’s hyperscale division hand-delivered the first units and described the delivery as the start of a broader commercial deployment.

Elon Musk reposted NVIDIA’s announcement on X and added that the chip “lived up to its name.” His public reaction followed the folding of xAI into SpaceX; the combined AI work now operates under the SpaceXAI name. NVIDIA’s AI Infrastructure account thanked SpaceX for testing the Vera units.

The Vera CPU contains 88 custom Olympus cores and supports up to 1.2 terabytes per second of memory bandwidth using LPDDR5X memory. NVIDIA reported that Vera runs agentic sandbox workloads up to 50% faster than competing rack-scale CPUs while doubling energy efficiency.

NVIDIA described a 256-CPU rack configuration that can sustain more than 22,500 concurrent agent environments at full performance. That rack is part of the Rubin platform the company unveiled at its GTC conference in March 2026. NVIDIA also said Vera already runs the Colossus 1 and Colossus 2 supercomputers in Memphis.

Customers listed as early adopters include Alibaba Cloud, ByteDance, Meta, CoreWeave, Lambda and Nscale, in addition to the initial recipients. NVIDIA noted several rivals, including AMD, are developing chips aimed at similar workloads. The company did not provide a timeline for large-scale commercial availability beyond the initial deliveries and testing.