Google’s October 1, 2026 launch of four Trillium TPUs aboard its Project Suncatcher MVP satellite on a SpaceX rocket marks the latest hardware validation in the orbital AI compute race, following Starcloud’s 2025 H100 GPU flight that trained a large language model in low Earth orbit. SpaceX has filed for up to one million data-center satellites and outlined 2027 production of its AI1 vehicle with roughly 120 kW of Nvidia Vera Rubin-class compute, while Starcloud secured $250 million to scale toward constellations of 88,000 satellites. These developments underpin the roughly 25% market-implied odds for a qualifying orbital data center—at least 100 GPUs or TPUs—by end-2027. Persistent challenges include radiative cooling in vacuum, inter-satellite laser links, radiation hardening, and launch economics, with Google’s 2027 follow-on satellites and potential Starcloud or SpaceX flights serving as near-term catalysts that could shift trader consensus.