Space-based data centers are emerging as a new frontier in artificial intelligence infrastructure. Companies in the United States, China, Canada and Europe are testing ways to move computing, data storage and AI processing beyond Earth.
The projects range from small satellites carrying AI processors to ambitious proposals involving tens of thousands or even one million orbital satellites. While some systems have already demonstrated computing capabilities in space, large-scale orbital data centers remain under development.
1. Operational and Flown Space Computing Projects
Several companies and research groups have moved beyond concepts and launched experimental hardware.
- Kepler Communications (Canada): Launched 10 satellites in January 2026 equipped with NVIDIA Jetson Orin multi-GPU modules, storage and optical inter-satellite links. Its commercial orbital computing and optical relay service opened to customers in 2026, with further expansion planned.
- China’s Three-Body Computing Constellation: Developed by ADA Space, Zhejiang Lab and partners, this network began with 12 satellites launched in May 2025. Its satellites support hundreds of TOPS of computing, onboard AI models of approximately eight billion parameters and optical links reaching 100 Gbps. Expansion plans range from hundreds to approximately 1,000–2,800 satellites.
- Starcloud (US): Its Starcloud-1 satellite, launched in November 2025, carried an NVIDIA H100 GPU and demonstrated AI workloads, including Google’s Gemma model. Starcloud is developing larger satellites and eventually gigawatt-scale orbital data centers. It has also pursued regulatory filings for tens of thousands of satellites.
- Google Project Suncatcher: Launched on October 1, 2026, with Planet-built satellite hardware and Google TPUs, the prototype successfully ran Gemma 4 in orbit and transmitted an AI-generated haiku to Earth. Google plans to test two interconnected satellites in 2027.
- Lonestar Data Holdings: Focuses on secure, resilient and sovereign data storage beyond Earth. Its projects include ISS and lunar payloads, the Freedom data centre concept and its planned StarVault orbital storage service, developed with Sidus Space. It also has a NASA Space Act Agreement.
- Axiom Space: Is developing orbital data-centre nodes and computing units connected through optical networks, including hardware on the International Space Station and free-flying platforms.
Other flown or experimental initiatives include Aethero, HPE Spaceborne Computer, Unibap, Voyager and Chinese satellite-computing demonstrations. China’s BUPT Tiansuan programme is also pursuing multi-satellite orbital computing.
2. Major Planned Orbital Data Center Constellations
Some of the largest proposals involve satellite networks designed specifically for AI computing.
| Company / Project | Proposed scale or objective |
|---|---|
| SpaceX / xAI | Up to 1 million AI satellites; potential 150–250 kW-class nodes and tens to over 100 GW of long-term capacity |
| Blue Origin Project Sunrise | Up to 51,600 satellites with optical networking |
| Orbital Compute | Up to 100,000 satellites, with approximately 10 GW potential capacity; pathfinder targeted for 2027 |
| Starcloud | Additional large-scale constellation filings |
| Cowboy Space | Approximately 20,000 satellites under its Stampede proposal |
| Sophia Space | Modular TILE architecture using solar power and passive cooling; initial satellites targeted for 2027–2028 |
| Aetherflux | Space-based solar-power constellation incorporating orbital computing nodes |
SpaceX’s proposed AI1 or Star Mind network is among the most ambitious concepts. Its plans involve dedicated satellite manufacturing and launches during the later 2020s.
Other emerging companies include OrbitsEdge, LEOcloud, SkyServe, Madari Space (UAE) and Rotonium, which are exploring orbital computing access, edge AI and quantum-related technologies.
3. European and Institutional Research
The European Union-backed ASCEND feasibility study, led by Thales Alenia Space under Horizon Europe, examined modular orbital data centres powered by space-based solar energy.
It explored their potential for reducing emissions and supporting European data sovereignty. The study identified possible technical and economic pathways involving low-emission launch vehicles and robotic assembly, with a roadmap towards gigawatt-scale infrastructure around 2050.
NVIDIA is also developing space-oriented computing platforms and modules, including technology related to Vera Rubin, while partnering with several companies in this emerging sector.
4. Can Space Data Centers Replace Earth-Based Facilities?
Not yet. Most operational systems remain demonstrators or edge-computing platforms, ranging from watts to low kilowatts.
Scaling them to megawatt or gigawatt levels requires breakthroughs in radiative cooling, radiation-resistant hardware, launch costs, reliability, repairability and satellite networking.
The potential advantages include near-continuous solar energy in suitable orbits, reduced dependence on terrestrial land, water and electricity grids, and processing satellite data closer to where it is generated.
The orbital computing race is no longer limited to one company or country. With more than 50 tracked programmes globally and regulatory filings exceeding one million proposed satellites, space could become an important extension of the world’s AI infrastructure.
