Orbital compute is often described as if vacuum itself were a cooling system. That misunderstanding keeps returning because space looks cold from a distance.
Cold is not the same thing as easy heat rejection. A compute platform in orbit cannot rely on air movement, cheap water, or large terrestrial cooling loops. It has to radiate heat away through surface geometry, timing, and disciplined workload shaping.
That is why orbital facilities end up behaving like thermal governments as much as datacenters. Job admission, scheduling, and failover policy all start to revolve around when heat can be safely shed instead of when demand arrives.
The result is a strange trade. Orbit can reduce land cost, sovereign dependency, and some environmental constraints, but it also makes bad thermal planning immediately structural. A terrestrial campus can sometimes throw hardware and water at a mistake. An orbital platform usually has to live with its envelope.
This is why the winners in orbital compute are unlikely to be the operators with the biggest model alone. They will be the operators who treat heat as a first-class product limit rather than a mechanical afterthought.