Nvidia has supplied initial volumes of its first in-house CPU designed for artificial intelligence agents, Vera, to SpaceX.

On May 19 local time, blockchain outlet BeInCrypto reported that Nvidia delivered early Vera silicon directly to SpaceX as well as OpenAI, Anthropic and Oracle Cloud.

The shipment is drawing attention as Nvidia's first processor that expands its AI infrastructure business, previously centered on GPUs, into the CPU segment. Nvidia said its hyperscale division delivered the first batch directly and presented the supply as the starting point for broader commercial deployment.

Vera includes 88 Olympus cores designed by Nvidia. Based on LPDDR5X memory, it supports up to 1.2 terabytes per second of memory bandwidth. Nvidia said the chip runs up to 50 percent faster than competing rack-level CPUs on agent sandbox workloads and delivers about twice the efficiency. It also said that in a rack configuration binding 256 CPUs, it can sustain more than 22,500 concurrent agent environments at full performance.

Vera is part of the Rubin platform unveiled at GTC in March 2026. Nvidia has been making clear a push to go beyond GPUs, the core of its revenue, and bundle the full infrastructure needed for AI agent services into its platform. Nvidia CEO Jensen Huang (젠슨 황) has previously defined AI agent-based services as a multi-trillion-dollar opportunity for the company.

Market reaction was also swift. Elon Musk posted a short message on X, formerly Twitter, saying "Vera nice, Vera nice" after Nvidia's AI infrastructure account thanked him for SpaceX's participation in testing.

SpaceX's early adoption also ties in with a recent reorganization of its AI operations. After integrating xAI into the company, SpaceX is running a unified AI division under the SpaceXAI brand. The organization is already operating the Colossus1 and Colossus2 supercomputers in Memphis. How quickly SpaceX moves Vera from a test phase into a live operating environment is seen as a variable that will determine future results.

The early customer group was not limited to SpaceX. Alibaba Cloud, ByteDance, Meta, CoreWeave, Lambda and Nscale were also included among customers waiting for deployment. This shows Nvidia aiming to expand its influence beyond training chips into the execution and operations phase for agents.

Still, it remains to be seen whether this will translate into changes in CPU market share. Meaningful adoption is possible only if customers complete testing and shift into full production environments. Competition targeting the same market has already started. Several companies, including AMD, are preparing rival chips aimed at similar workloads. As a result, Vera's success is expected to be determined by the speed of actual large-scale operational transitions rather than early customer wins.

Thanks @SpaceX and @elonmusk, excited for you to try out the NVIDIA Vera CPU https://t.co/FYDFIXt33Y