Google plans to launch an experimental orbital AI computer called MVP on October 1 as the first hardware test for Project Suncatcher. The refrigerator-sized satellite, built and tested by Planet Labs, is due to fly from California on a SpaceX Falcon 9 and should have enough compute to answer basic AI queries.
The long-term concept is to run data-center hardware on abundant solar power in orbit. That could reduce pressure on terrestrial power grids, but it replaces familiar constraints with difficult space engineering. There is no air for conventional cooling, radiation can damage chips, and failed hardware is costly to repair or replace.
Scale is the largest limitation. A Google manager estimated that matching one 1-gigawatt ground data center would require roughly 10,000 satellites. Launch economics would also need to improve sharply, to around $200 per kilogram, before orbital compute becomes competitive. Jeff Bezos has estimated that cost parity may be as far as 20 years away.
SpaceX and Blue Origin are exploring similar concepts. The October mission is therefore a feasibility experiment, not the beginning of near-term migration for ordinary AI workloads. Its useful result will be evidence about power, cooling, radiation and communications in a real orbit.