SpaceX Partners with NVIDIA to Launch Vera Rubin AI Supercomputers into Orbit
SpaceX has announced a partnership with NVIDIA to power its upcoming Starmind AI1 satellite constellation using NVIDIA's next-generation Vera Rubin NVL72 architecture. This initiative aims to establish a distributed orbital data center capable of running heavy AI workloads directly in space. Bringing datacenter-class AI hardware to orbit could drastically reduce latency for space-based data processing and enable real-time, complex analysis without relying on ground stations. It represents a significant leap toward building space-based cloud computing infrastructure. The Vera Rubin NVL72 platform combines 36 Vera CPUs and 72 Rubin GPUs, delivering up to 3.6 exaFLOPS of AI performance. However, operating such high-power, liquid-cooled hardware in space presents massive challenges regarding thermal management, radiation hardening, and power supply.
## BACKGROUND
Starmind is SpaceX's planned satellite constellation designed to function as a distributed orbital supercomputer. NVIDIA's Vera Rubin NVL72 is an advanced rack-scale AI platform designed for agentic AI workloads, utilizing liquid cooling and high-speed NVLink interconnects. Traditionally, space-based computing has been limited to low-power, radiation-hardened processors due to the harsh environment of space.