Google’s October 1 launch of its Project Suncatcher prototype satellite carrying four TPUs aboard a SpaceX rocket marks the latest step in testing orbital AI compute, driven by surging terrestrial data center power demands and the appeal of continuous solar energy plus radiative cooling in vacuum. Starcloud’s 2025 satellite with a single Nvidia H100 GPU demonstrated early large language model inference in orbit, while SpaceX has unveiled its AI1 design targeting up to 150 kW payloads and filed for constellations of up to one million satellites, with deployments eyed for 2027–2028. However, qualifying as a data center under typical market criteria requires at least 100 accelerators plus successful launch, a threshold current tests fall far short of amid radiation hardening, inter-satellite laser links, and launch cost hurdles. Blue Origin and startups have filed similar large-scale plans, but credible timelines remain uncertain given historical slips in complex space hardware.