SpaceXAI adopts NVIDIA Vera CPUs for agentic AI infrastructure

NVIDIA says SpaceXAI will deploy NVIDIA Vera CPUs to accelerate the CPU-heavy work behind its next generation of agentic AI systems. The companies frame Vera as a processor built for the parts of an agent workload that sit around model inference: orchestration, tool use, code execution, data processing and simulation.

The announcement also ties SpaceXAI’s Grok infrastructure to NVIDIA’s Vera Rubin platform. NVIDIA says SpaceXAI is scaling toward gigawatt-class AI factories and plans to extend an optimized Vera Rubin NVL72 architecture into orbit through a first-generation Starmind AI satellite.

The important signal is not just another chip customer. NVIDIA is arguing that the next bottleneck for AI agents is full-system infrastructure: CPUs that keep tools and data moving, GPUs that handle model work, networking that keeps large clusters fed and software that makes the whole setup behave like an AI factory.

For AI users, this is another sign that agentic systems are moving from demos into expensive industrial infrastructure. The companies building them are now optimizing for long-running work, low-latency tool use and more useful output per watt, not only bigger model training runs.