SpaceX and Nvidia want to put Rubin GPUs and Vera CPUs in AI satellites

SpaceX and Nvidia are jointly developing the computing hardware for the Starmind AI1 satellite, which will carry Nvidia's Rubin GPUs and Vera CPUs on board. With this, the two companies are taking a step toward offering AI computing capacity from orbit.

Elon Musk, CEO of SpaceX, announced the collaboration via his social network X. He indicated that SpaceX will exclusively use Nvidia architecture for its future AI infrastructure. Jensen Huang, founder and CEO of Nvidia, previously introduced the Vera Rubin architecture and the accompanying Space-1 Vera Rubin module, which is specifically designed for use in space.

The announcement fits into a broader series of partnerships that Nvidia is forming with space companies to bring accelerated computing capacity beyond Earth's atmosphere.

What the Starmind AI1 satellite is intended to be

The Starmind AI1 satellite is the concrete project for which SpaceX and Nvidia are jointly designing the so-called 'compute payload', the part of the satellite that performs the actual computing work. By combining Rubin GPUs and Vera CPUs, SpaceX aims to be able to run AI models and compute-intensive processes directly in space, rather than sending data to ground stations for processing.

The Vera Rubin architecture is Nvidia's next-generation chip architecture, named after American astronomer Vera Rubin. Jensen Huang presented the Space-1 Vera Rubin module as a variant capable of handling the specific conditions in space, such as radiation and thermal extremes. Based on available source material, the launch date of the first Starmind satellite has not yet been confirmed.

The combination of GPU and CPU power in a satellite makes it possible in principle to perform tasks that would normally require a data center on Earth, such as processing large volumes of sensor data or running AI inference close to the source of the data.

Nvidia's position in the space sector

The collaboration with SpaceX is not Nvidia's only move toward the space market. The company is also working with Axiom Space, Cowboy Space Corporation, Firefly Aerospace, Kepler Communications, Muon Space, Sophia Space, Starcloud, and Aetherflux on the deployment of accelerated computing capacity in satellites and space platforms.

For the Vera CPUs, Nvidia counts early customers including OpenAI, Oracle, and Anthropic alongside SpaceX. That broader customer list shows that Vera is not limited to space applications, but is becoming part of Nvidia's standard data center offering. The space market remains a niche segment for Nvidia for now, but the partnership with SpaceX gives that market a considerably higher profile.

Musk's statement that SpaceX will use exclusively Nvidia architecture for future AI infrastructure is notable, particularly given his earlier involvement with xAI, which is building its own infrastructure. Whether that statement extends to the full AI activities of his companies or refers specifically to the satellite hardware cannot be fully determined based on the available information.

Why AI data centers in space are relevant

The concept of computing capacity in satellites is not entirely new, but the combination with modern AI hardware has been drawing increasing attention in recent months. Conventional satellites send raw data to the ground for analysis. If that analysis can take place on board, it reduces the required bandwidth and shortens the time between measurement and decision. This is relevant for applications such as Earth observation, defense, and connectivity in remote areas.

At the same time, computing capacity in space brings its own challenges. Chips must be resistant to cosmic radiation, thermal management is more complex than in a climate-controlled data center, and the power supply via solar panels is more limited. According to Nvidia, the Space-1 Vera Rubin module is tailored to those conditions, but how its performance compares to ground-based hardware in practice has not yet been made public.

For the Dutch and European space and AI sectors, announcements like this signal that major technology companies are increasingly viewing space as an extension of their cloud infrastructure. European parties such as the European Space Agency and companies active in Earth observation or satellite communications are thus faced with the question of how far they should seek alignment with American platforms or build their own capacity.