As demand for computing power grows faster than the grid supplying it, proposals to move data centres into orbit have moved from speculation to funded projects.
The case is not absurd. It is also considerably harder than the pitch decks suggest.
The arguments in favour
Solar power in orbit is continuous and about forty percent stronger than at the surface, with no night in the right orbit and no weather. That is genuinely abundant energy in a place where nobody has to be persuaded to allow a substation.
There is no land use, no local objection, no competition with residential electricity demand, and no water consumption for cooling.
The cooling problem
This is where it gets hard. On Earth, heat is removed by moving air or water past hot components. In vacuum there is nothing to move it into. The only mechanism available is radiation, which is far less efficient per unit of area.
A facility drawing serious power needs radiator panels of substantial size – large, fragile structures that add mass, cost, and something else to fail.
The other obstacles
Radiation degrades electronics. Hardened components exist, are generations behind the current state of the art, and cost far more. Running consumer-grade hardware in orbit means accepting a failure rate.
Maintenance is the deeper issue. In a terrestrial data centre a failed drive is replaced by someone walking to the rack. In orbit there is nobody, and the economics depend entirely on how long hardware lasts without intervention.
And latency rules out anything interactive. A user waiting on a response cannot wait for a signal to reach orbit and return, which limits the use case to batch work.
An honest assessment
Plausible for specific workloads that are compute-heavy, latency-tolerant, and can run untouched for years. Not plausible as a replacement for terrestrial infrastructure, and the launch costs only work if hardware lifetime in orbit turns out much better than current experience suggests.