SpaceXAI Deploys NVIDIA Vera CPUs to Power Gigawatt-Scale Agentic Infrastructure

SpaceXAI is deploying NVIDIA's Vera CPUs to handle the intensive processor workloads of its next-generation agentic AI applications. [1] The company's AI infrastructure buildout supporting its Grok models is currently scaling toward gigawatts of computing capacity. [1] Agentic systems rely heavily on conventional processors to execute code, process data, and orchestrate tools between model calls. [1] When these CPUs lag, the GPUs they support sit idle, which is the bottleneck the Vera processor was built to address. [1]

Vera is NVIDIA's first CPU built specifically for agentic AI, featuring 88 custom Olympus cores and Spatial Multithreading technology. [1] The processor utilizes LPDDR5X memory to provide up to 1.2 terabytes per second of bandwidth, which NVIDIA says completes tasks up to 1.8 times faster than x86 CPUs in reinforcement learning and agentic AI workloads. [1] The deal will also expand into space, with SpaceXAI planning to use an optimized NVIDIA Vera Rubin NVL72 rack-scale system for its first-generation Starmind AI satellite. [1]

The deployment signals a structural shift in AI infrastructure priorities, highlighting that as models move from generating text to executing autonomous actions, CPU bottlenecks become a critical limit on GPU efficiency. By designing the Vera CPU specifically for agentic workloads, NVIDIA is attempting to control the entire compute pipeline and challenge traditional x86 dominance in orchestration tasks. Furthermore, putting a Rubin NVL72 system into orbit suggests highly ambitious plans for space-based, rack-scale edge computing. What the sources don't address: How the immense power and cooling requirements of the Vera Rubin NVL72 architecture will be practically managed in an orbital satellite environment.