SpaceX plans Starmind satellites with powerful AI computing hardware and solar arrays.
NVIDIA Vera Rubin systems could deliver powerful AI computing capabilities in orbit.
Orbital data centers could reduce energy, latency and terrestrial infrastructure pressures.
The race to build bigger and more powerful artificial intelligence systems is moving beyond Earth. SpaceX and NVIDIA are developing orbital AI infrastructure that could eventually turn satellites into data centers, bringing computing closer to space-based data sources and tapping solar power.
SpaceX is developing its Starmind AI satellite around NVIDIA’s Vera Rubin computing architecture. The project marks a significant step towards putting high-performance AI computing infrastructure in orbit.
SpaceX plans to deploy AI satellites equipped with powerful computing hardware and solar arrays. Its Starmind AI1 design is intended to perform AI processing in orbit and transmit results to Earth through high-bandwidth laser links connected to the Starlink network.
The satellite is designed around an optimized NVIDIA Vera Rubin NVL72 system. NVIDIA said the Starmind platform is being adapted to meet the different power, thermal, bandwidth, and reliability requirements of operating in space.
According to SpaceX’s current specifications, the AI1 satellite can deliver up to 250 kW of peak compute power and 175 kW of average power.
The approach could allow computing workloads to be handled directly in orbit rather than relying entirely on ground-based infrastructure.
One of the central arguments behind orbital computing is access to energy.
AI data centers on Earth require enormous amounts of electricity. As demand for AI computing grows, companies face increasing pressure to secure sufficient power, cooling infrastructure, and grid capacity.
Satellites operating in orbit can use solar energy without depending on terrestrial power grids. SpaceX argues that access to abundant solar power could help provide a path to scaling AI infrastructure as demand continues to rise.
Orbital computing could also address another challenge: the huge amount of information collected by satellites.
Satellites generate vast quantities of imagery and sensor data. Instead of sending all raw data back to Earth for processing, orbital computers could analyze the information onboard and transmit only the results that matter. NVIDIA said this approach can reduce downlink requirements and latency.
For NVIDIA, space-based computing represents an expansion of its AI infrastructure business beyond traditional data centers.
The company is developing its Space-1 Vera Rubin module, along with Jetson Orin and IGX Thor platforms, for environments where size, weight, and power are tightly constrained.
NVIDIA said Space-1 can deliver up to 25 times more AI compute per GPU for space-based workloads.
The Starmind project also allows NVIDIA to extend a common hardware-and-software architecture from terrestrial AI factories into orbit.
Also Read: NVIDIA Joins Nscale’s $2B Round as AI Data Centre Demand Grows
The technology faces major challenges. Launching, maintaining, and replacing computing hardware in orbit remains expensive. Space-based electronics must also withstand radiation and operate under demanding thermal and reliability conditions.
Unlike an Earth-based data center, an orbital facility cannot simply be accessed by engineers when a component fails. Hardware must be designed to withstand the environment, while repairs and replacements pose additional logistical challenges.
SpaceX is betting that Starship’s eventual high launch cadence and reusability can dramatically lower the cost of putting infrastructure into orbit.
The company is also planning a major expansion of its launch infrastructure. A proposed $100 billion Starbase Louisiana project is designed to support future Starship operations and satellite deployment.
SpaceX and NVIDIA are targeting the first-generation Starmind AI satellite for 2027, with SpaceX aiming for significant expansion in 2028.
The ambition extends far beyond a small number of satellites. SpaceX is pursuing plans that could eventually involve up to one million AI satellites, potentially creating a large-scale orbital computing network.
Whether such a system can become economically viable remains uncertain. The cost of launches, hardware reliability, maintenance, and orbital operations will determine how quickly the concept can scale.
For now, the SpaceX-NVIDIA partnership points to a broader shift in the AI infrastructure race. As demand for computing power continues to rise, the next generation of AI infrastructure may not remain confined to land-based data centers. Increasingly, the industry is looking towards orbit as another potential home for computing.
Why this MattersSpaceX and NVIDIA’s orbital AI push could reshape computing by combining abundant solar power, onboard data processing and large-scale satellite infrastructure. If successful, space-based data centers could reduce terrestrial energy pressures while creating a new frontier for AI computing and infrastructure.
What is SpaceX's Starmind AI satellite?
Starmind AI1 is a SpaceX satellite designed to process artificial intelligence workloads in orbit using advanced NVIDIA computing hardware.
Why is SpaceX developing AI data centers in space?
SpaceX sees orbital computing as a way to use solar power while reducing pressure on Earth-based electricity and infrastructure.
What NVIDIA technology will Starmind use?
Starmind AI1 will use an optimized NVIDIA Vera Rubin NVL72 system adapted for space's demanding computing environment.
When will SpaceX launch its first Starmind AI satellite?
SpaceX and NVIDIA are targeting 2027 for the first-generation Starmind AI satellite, with significant expansion planned for 2028.
How many AI satellites could SpaceX eventually deploy?
SpaceX is pursuing plans that could eventually involve up to one million AI satellites forming a large-scale orbital computing network.