Orbital AI Data Center Glossary
An orbital AI data centre is a satellite or orbital platform designed to perform substantial AI computation in space rather than sending every raw input to a terrestrial data centre.
The idea moved beyond a thought experiment in 2025–26. Start-ups pursued in-orbit compute, and SpaceX filed plans with the U.S. Federal Communications Commission for a very large constellation of “orbital data centre” satellites. A filing is evidence of intent, not evidence that the proposed scale is technically or economically achievable.
The attraction is clear enough: abundant solar energy, no terrestrial land use, and the possibility of processing some satellite data before transmitting results to Earth. In the SpaceX/xAI case, the proposal also joins launch, satellite networking, frontier AI, and financing inside one corporate system.
The caution is equally clear. Space does not abolish physics. It changes the physics bill. Launch mass, radiation hardening, maintenance, latency, orbital debris, heat rejection, capital cost, insurance, geopolitics, and failure recovery all remain. Anyone selling orbital compute as escape from constraint is selling a fairy tale with better rockets.
For this Dictionary, the phrase is both an engineering proposal and a symptom of the AI buildout’s ambition. Its near-term plausibility depends on workload, launch cost, power density, communications, reliability, and thermal design—not on the romance of putting a data centre above the weather.
See also
Sources
- Data Center Dynamics, “SpaceX files for million-satellite orbital AI data center megaconstellation”, 2 February 2026.
- CNBC, “No one wants AI data centers on Earth. Do they make sense in space?”, 21 June 2026.