Nvidia has unveiled the Vera Rubin Space Module, a new AI compute platform engineered specifically for orbital data centers, promising up to 25 times the AI processing capability of its H100 GPU. Announced during the GTC 2026 event, this ambitious hardware aims to power advanced large language models and foundation models directly in space, shifting AI workloads closer to the data source rather than relying on ground stations.
The Vera Rubin Space Module features a tightly integrated CPU-GPU architecture with a high-bandwidth interconnect. This design addresses the unique challenges of managing massive, continuous data streams from space-based instruments, enabling real-time inference and analysis aboard satellites and orbital platforms. Nvidia targets applications that require on-the-spot data processing, such as Earth observation, satellite communications, and space science.
Six commercial space companies are reportedly already deploying Nvidia’s platform in orbit or on the ground: Aetherflux, Axiom Space, Kepler Communications, Planet Labs PBC, Sophia Space, and Starcloud. Kepler Communications, for example, uses Nvidia’s Jetson Orin modules across its satellite constellation to perform AI-driven data management in space. ”Nvidia Jetson Orin brings advanced AI directly to our satellites, allowing us to intelligently manage and route data across our constellation,” said Mina Mitry, Kepler’s CEO.
This push for space-based AI reflects a broader shift across the industry, where reducing latency and bandwidth usage by processing data near its source-in satellites rather than on Earth-offers vast benefits. It’s also a leap from traditional space hardware to systems capable of running foundational AI models, suggesting future satellites might autonomously adapt and optimize without constant ground control intervention.
The Vera Rubin Space Module does not yet have a release date, with Nvidia only stating that availability will come ”at a later date.” Meanwhile, other Nvidia space-focused products, such as the Jetson Orin, IGX Thor, and RTX PRO 6000 Blackwell Server Edition, remain actively supplied for both orbital and terrestrial AI workloads.
Key features of the Nvidia Vera Rubin Space Module include:
- Up to 25 times the AI processing power of the Nvidia H100 GPU
- Tightly integrated CPU-GPU architecture with high-bandwidth interconnect
- Designed for real-time inference and analysis aboard satellites and orbital platforms
- Optimized for applications like Earth observation, satellite communications, and space science
- Supports deployment of large language and foundation AI models in space
In-orbit computing is ramping up fast, and Nvidia’s commitment signals that satellite operators want high-performance AI capabilities to handle both inferencing and data orchestration in space. This approach can potentially ease pressure on ground infrastructure and networks. The Vera Rubin Space Module could redefine how space and AI intersect, especially as satellite constellations grow larger and data volumes explode.