NVIDIA Conquers Space: Launch of Orbital Data Center with H100 GPU
NVIDIA, a leader in GPU production, announced the launch of an ambitious project to create orbital data centers. In partnership with startup Starcloud, it is preparing to launch the Starcloud-1 satellite, equipped with the H100 GPU, to test AI computing in space. This initiative aligns with the interest of figures like Jeff Bezos, Demis Hassabis, and Sam Altman in space computing platforms.
Launch and Technical Details
On October 23, 2025, NVIDIA posted information on its blog about the collaboration with Starcloud. The 60-kg Starcloud-1 satellite will be launched into orbit by a SpaceX Falcon 9 rocket in November 2025. On board is the flagship H100 processor, intended for experiments with AI under space conditions. This is the first time NVIDIA equipment will be tested in an orbital environment, marking the beginning of the era of space data centers.
Crusoe, another partner, plans to integrate its cloud services into this infrastructure. From late 2026, Crusoe will start providing limited GPU capacity, and by 2027 a public cloud based on satellite computing.
Project Goals and Benefits
The project is aimed at creating orbital data centers fully powered by solar energy. Future satellites are planned to be equipped with solar panels up to 4 km wide, utilizing solar radiation without dependence on Earth's energy sources. Space vacuum is positioned as an “infinite radiator” for cooling, eliminating the need for water systems and reducing the environmental footprint.
NVIDIA claims that orbital data centers will allow for a tenfold reduction in energy costs compared to terrestrial counterparts, even considering launch expenses. By 2030, the company plans to deploy gigawatt-scale orbital clusters for processing global AI tasks such as modeling, data analysis, and neural network training.
Potential for the Crypto Industry
Orbital data centers could change the approach to cryptocurrency mining and blockchain applications. Mining requires significant energy consumption and efficient cooling, often causing environmental concerns. Space data centers, operating on solar energy and not using water, could lower these barriers. Integration of AI to optimize mining algorithms, such as proof-of-work or proof-of-stake, could increase efficiency, though NVIDIA has not yet disclosed plans to adapt the technology for the cryptocurrency industry.
Criticism and Technical Challenges
Experts express skepticism about the project. Firstly, the vacuum of space is an insulator, not an ideal heat sink. To dissipate heat, large infrared radiators are required, which increase the vulnerability of satellites to micrometeoroids and reduce efficiency compared to terrestrial systems.
Secondly, H100 GPUs are susceptible to cosmic ray effects, causing computational errors. Shielding is required for protection, which increases the mass and cost of the satellite. Data transmission from orbit to Earth is accompanied by millisecond delays, making it impossible for real-time AI tasks, limiting applications to offline processing, such as satellite image analysis.
The environmental aspect also raises questions: the launch of a Falcon 9 rocket is accompanied by significant CO2 emissions, which may exceed the savings from data center operation. Launch costs, including the $30,000 H100 and the satellite itself, reach millions of dollars. Satellite maintenance is impossible, and their lifespan (5+ years) exceeds the GPU upgrade cycle (2–3 years), reducing economic viability.