Intelligence must live wherever data is generated. The convergence of space-based compute and terrestrial AI infrastructure is no longer science fiction.
The dominant story in AI infrastructure for the past three years has been concentration: more compute, denser packaging, larger pods, gigawatt sites. That story is real, but it’s not the whole story.
Edge silicon shipped $278 billion in 2025 and is on a path to $340 billion by 2030. That trajectory is being driven by a much simpler observation: a growing number of AI workloads cannot move the data they need to act on, either because of physics (latency), economics (network egress), or sovereignty (regulation).
Robotics, autonomous vehicles, defense platforms, industrial controls, and increasingly, orbital sensors are all generating intelligence at the edge — and using purpose-built silicon to make decisions before the data ever sees a datacenter.
Starcloud and similar orbital infrastructure startups are now extending that pattern into space, where solar power is essentially free and cooling is essentially free. It’s not science fiction. The economic case for moving more inference off the ground is real, and the silicon stack to do it is shipping today.
