Nvidia Launches Space-Grade Vera Rubin GPU for Orbital AI
Nvidia built a space-specific version of its Vera Rubin platform that delivers 25x more AI compute than H100 in orbit.
Nvidia just took its AI hardware off-planet. The company unveiled Space-1 Vera Rubin, a space-specific module of its Vera Rubin GPU-CPU platform designed explicitly for orbital data centers.
The numbers are wild. Nvidia claims the chip delivers up to 25x more AI inferencing performance compared to the H100 — in space. That's a massive leap for satellite-based computing.
Six companies are already lined up as partners: Aetherflux, Axiom Space, Kepler Comms, Planet, Sophia Space, and Starcloud. The customer list spans satellite operators, space station builders, and orbital communications providers.
This is Nvidia's clearest bet yet that AI workloads won't stay earthbound. Processing data where it's collected — in orbit — eliminates the latency and bandwidth costs of beaming everything back down. Space-based inferencing just got serious silicon behind it.