SpaceX to Deploy 10GW AI Compute Using NVIDIA GPUs
Elon Musk confirms SpaceX will exclusively use NVIDIA's Vera Rubin platform for global and orbital AI data centers through 2027.
SpaceX Founder and CEO Elon Musk has announced a commitment to NVIDIA GPUs for all company AI requirements, utilizing the Vera Rubin platform. The deployment includes the Kyber-codenamed NVL72 rack and the Space-certified Space-1 Vera Rubin module, which features four Rubin GPUs and two Vera CPUs. SpaceX plans to reach 2GW of compute capacity by late 2026, scaling to 10GW by the end of 2027. In a partnership with Anthropic, SpaceX is developing an Orbital AI data center to bypass terrestrial constraints such as land availability, power access, and cooling requirements. These AI workloads will be distributed across both terrestrial and space-based installations to optimize performance and overcome traditional infrastructure bottlenecks.