Orbital raises $5M pre-seed to build AI data centres in space
What's the deal? Los Angeles-based startup OrbitalDealroom has a profile for this one. Try Dealroom → has raised an oversubscribed $5M pre-seed round to build AI data centres in low Earth orbit. The company is designing satellites purpose-built for AI inference — the fastest-growing segment of compute demand — that harness continuous solar energy and radiative cooling directly into space.
Each satellite is engineered for 100 kW of compute power. The long-term vision: a constellation of more than 100,000 satellites delivering over 10 GW of orbital compute.
Why now? AI inference demand is surging, and terrestrial data centres face growing constraints around energy, cooling, and land. Orbital argues that space solves all three at once — uninterrupted solar power, natural radiative cooling, and no need for real estate.
The company's systems are being designed around Nvidia's publicly announced Space-1 Vera Rubin-class GPU architecture. Its first mission, Pathfinder, is slated for 2027 and will fly a hosted GPU payload on a SpaceX Falcon 9 rideshare to test GPU operation, radiation tolerance, thermal performance, and data downlink in orbit. A purpose-built satellite, Orbital-1, will follow.
To support production at scale, Orbital is developing Factory-1, a satellite assembly and testing facility in the South Bay area of Los Angeles.
What could go wrong? The technical hurdles are enormous. Operating GPUs in space means contending with radiation, extreme thermal swings, and the challenge of high-bandwidth data downlink back to Earth. Orbital's chief executive, Poon, acknowledged the company is "tackling some of the hardest engineering challenges in the industry" at a satellite size and constellation scale never attempted before.
A constellation of 100,000+ satellites is an extraordinarily ambitious target that would dwarf even SpaceX's Starlink network. Manufacturing, launching, and maintaining that fleet would require billions in capital and years of execution — all from a company still at the pre-seed stage.
The signal: Orbital sits at the intersection of two powerful trends: the insatiable demand for AI compute and the rapid decline in launch costs that has made ambitious space infrastructure viable. If terrestrial data centres keep hitting power and cooling bottlenecks, space-based alternatives could shift from science fiction to serious infrastructure.
But the gap between a $5M pre-seed and a 10 GW orbital compute network is vast. Pathfinder in 2027 will be the first real proof point — and investors will be watching closely.
Read more: Benzinga
