Google Sends TPUs to Orbit for First Time in Suncatcher Test
By @sharedot · · 6 pages
- Tech
- Space
- AI Hardware
Google's Project Suncatcher prototype carried TPU AI chips into orbit on SpaceX's Transporter-18 to test if AI computing can survive in space.
What Happened: TPUs Reach Orbit for the First Time
Alphabet has put its Tensor Processing Units into orbit for the first time, launching the first Project Suncatcher prototype aboard SpaceX's Transporter-18 rideshare mission from Vandenberg Space Force Base in California. SpaceX confirmed the rideshare carried 130 payloads to low-Earth orbit.
The satellite platform was built with Planet Labs, which Google enlisted to test its TPUs in space and eventually demonstrate high-bandwidth communication between satellites.
Planet reported initial contact with Tanager-2, the Suncatcher prototype, and two SuperDoves, its 40th successful launch and 718th satellite delivered to orbit.
Why It's a Big Deal: AI Compute Meets Space Power
This is not an orbital data center yet, but it is the first real hardware step toward one.
Google says satellites in favorable low-Earth orbits could receive up to eight times more solar power than ground-based panels, with much longer access to sunlight — a direct answer to the power constraint that limits terrestrial AI data centers.
Google's research previously estimated that if launch costs fall to around $200 per kilogram, the economics of space-based computing could begin to approach those of data centers on Earth. As Aaron Burnett, CEO of space-tech research firm Mach33 Financial Group, put it: '1kw of TPUs per sat.
Godspeed to Google... Now they just need to scale that up 100x power.'
The Evidence: Radiation and Cooling Under Test
The mission is designed to determine whether the chips can withstand launch forces, radiation and the unusual thermal environment of space.
According to Stocktwits, tests of Trillium TPUs with a proton beam showed the chips could survive radiation beyond a five-year mission, though the attached high-bandwidth memory proved more sensitive.
TradeArabia reports the cooling system uses heat pipes moving heat to radiators, with the chips running only about 15 minutes at a time before shutting down so radiators can catch up. Google will also run Gemini models on the TPUs during the mission to evaluate their performance in orbit.
What's Next: Three Problems Before Scale
Radiation, cooling and networking remain the key challenges Google must solve before any large-scale orbital AI infrastructure becomes real. Longer-term, the vision calls for satellite clusters linked by high-bandwidth optical connections rather than a single spacecraft.
Results from this prototype will be used to identify points of failure and refine designs for future missions, TradeArabia reports.
The announcement lands amid mixed market sentiment: Stocktwits notes GOOGL fell 4% on Wednesday and another 1% Thursday, pressured partly by a judge letting a $3.2 billion ad-tech class action proceed, even as retail sentiment on the platform moved 'bullish.'