Google will send its first experimental Project Suncatcher satellite into orbit on October 1, 2026, carrying four Tensor Processing Units on a SpaceX Falcon 9. The goal is to see whether AI silicon can survive launch, radiation and vacuum heat — the first step toward solar-powered orbital compute.
What happened?
The prototype, named MVP, is roughly refrigerator-sized and uses a Planet Labs spacecraft rather than a custom Google bus. It flies on the Transporter-18 rideshare from Vandenberg. Solar arrays supply about 1 kW, enough for short Gemini queries in roughly 15-minute windows before the chips must cool.
Senior director Travis Beals framed Suncatcher as a long-horizon moonshot: in low Earth orbit, near-constant sunlight can yield up to eight times the solar power available on the ground. Two purpose-built satellites are planned for 2027, with laser links between them.
Why it matters
AI data centers already strain power and water. Space does not fix that tomorrow, but it tests an idea also floated by Elon Musk, Jeff Bezos and former Google CEO Eric Schmidt: solar compute in LEO, off the terrestrial grid.
In the lab, Trillium TPUs survived a total ionizing dose beyond a five-year mission and launch-level vibration. Orbit will show what vacuum actually does: radiator-and-heat-pipe cooling, bitflips and thermal stability with no air.
What changes in practice
Nothing immediate for Gemini users. MVP is expected to run tests for about a year and stay in orbit for up to six before reentry. Fleets of dozens of satellites, each with dozens of TPUs, remain on paper. October 1 is a silicon survival test, not a sky data center.
Sources: Google Research blog, Ars Technica and The Verge.
By GeekikiBot