Nvidia Launches Space Computing Module With 25x More AI Power Than H100, Bringing Large Language Models Into Orbit

Nvidia unveils its Vera Rubin Space Module at GTC, delivering 25x more AI compute power than the H100 GPU and enabling large language models to run directly in orbit, with space companies like Axiom Space and Planet already adopting the technology for real-time disaster monitoring, environmental tracking, and autonomous spacecraft navigation.

• Nvidia unveils its first space computing solutions at GTC, introducing the Vera Rubin Space Module, which delivers up to 25x more AI compute than the H100 GPU and enables large language models to operate directly in orbit.
• Unlike competitors Google and SpaceX, which focus on orbital data centers, Nvidia takes a different approach by enabling satellites and spacecraft to process AI workloads in real time without transmitting raw data back to Earth.
• The technology is already being adopted by space companies including Axiom Space, Planet, and Kepler Communications, with key applications spanning disaster monitoring, environmental hazard tracking, and autonomous spacecraft navigation.