We build solar-powered satellites that run AI inference in continuously sunlit low Earth orbit, beyond terrestrial power, land, and water limits.
Steady Light Compute designs spacecraft that function as distributed AI compute nodes in low Earth orbit. Each satellite carries its own solar arrays and rejects heat by radiating into space, so the payload can run inference workloads without drawing on a ground grid.
Workloads are split across many independently deployable satellites rather than one tightly coupled cluster. That architecture fits AI inference, which can run node by node instead of requiring a single synchronized training fabric.
The goal is commercial AI compute infrastructure in low Earth orbit for organizations that need capacity the terrestrial grid cannot reasonably supply.
A first hosted GPU demonstration flight is planned within about two years to validate GPU operation, radiation tolerance, thermal performance, and data downlink in orbit. A first purpose-built compute satellite, with multiple GPU nodes and high-bandwidth ground links, is planned to follow. Production satellites are being designed for high tens of kilowatts of compute power per spacecraft, with a longer-term vision of a distributed constellation.
Each spacecraft generates its own power from a continuously sunlit orbit, reducing dependence on terrestrial electricity and permitting timelines.
Heat is rejected passively into space, avoiding the water and site constraints that limit ground data centers.
AI inference workloads can be partitioned across many independently deployable satellites instead of one tightly coupled cluster.
Systems are engineered for radiation, thermal performance, and data downlink in low Earth orbit.
Build commercial AI compute infrastructure in low Earth orbit so inference workloads can scale without drawing on terrestrial electricity, land, or water.