SpaceX wants to build AI data centers in space. Will it work?
TL;DR
SpaceX and others are exploring orbital data centers as AI pushes demand for power, land and cooling. The pitch is abundant solar power, less water use and fewer local fights over grid load, land use and noise. The main problem is operations, not launch. Servers turn most electricity into heat, and in vacuum there is no air to carry it away. Radiators for 10 megawatts of waste heat could need surfaces roughly the size of two football fields.
Nauti's Take
This is not cloud computing moving to space overnight; it is a PR-heavy infrastructure bet with serious engineering debt attached. SpaceX has real advantages in launch and satellite operations, but servers are not Starlink terminals: they age fast, fail, need cooling and have to survive radiation.
The sensible first market is data already created in orbit. For latency-sensitive AI services on Earth, terrestrial data centers remain messy but much more practical.
Briefingshow
AI compute is becoming an infrastructure problem: power, cooling, water use, latency and local acceptance all matter. Orbit reduces some Earth-side constraints, but turns them into harsh engineering problems with high cost and little tolerance for failure. The first real market is likely space-generated data, not ChatGPT in orbit.
Sources