Google is launching its AI chips into space — Project Suncatcher MVP sports four TPUs and will be delivered by a SpaceX rocket
- Google is testing whether AI hardware can survive repeated exposure to space
- The experimental satellite carries four custom Tensor Processing Units into orbit
- Radiation testing showed the TPUs could withstand demanding orbital exposure levels
Google will launch its first experimental orbital data center satellite on October 1 2026, marking an early test for Project Suncatcher, its long-term research initiative.
This moonshot effort explores whether low Earth orbit could eventually host scalable machine learning infrastructure powered largely by sunlight.
The satellite, called MVP, carries four of Google’s custom Tensor Processing Units inside a spacecraft roughly the size of a refrigerator.
Latest Videos FromTechRadar
Testing hardware survival before scaling any further
A SpaceX Falcon 9 rocket will carry the satellite into orbit as part of the Transporter-18 rideshare mission alongside other payloads.
Satellites in low Earth orbit can generate up to eight times more solar power than equivalent panels positioned on Earth’s surface.
Google did not build MVP from scratch; instead, it integrated its chips into a satellite that Planet Labs, a satellite imagery company, had already built, helping Google accelerate its original 2027 timeline for launching two custom-built satellites into orbit.
During a rocket launch, spacecraft experience vibration and acceleration loads reaching up to ten times normal gravity throughout the flight.
Individual chip components can experience even greater forces, ranging between fifty and one hundred times normal gravity during ascent.
Engineers tested the satellite by shaking it along three separate axes to simulate those exact launch conditions beforehand, and that vibration testing confirmed the hardware withstood the simulated launch forces without any structural or functional failure.
“Tests like this rarely go as planned, so we were pleasantly surprised that the hardware held up to the force,” said Travis Beals, Senior Director, Paradigms of Intelligence at Google.
Radiation exposure presented another concern, so Google tested its Trillium TPUs inside a proton beam facility at UC Davis.
Those tests showed the chips survived radiation doses exceeding what a five-year orbital mission would typically deliver.
Cooling remains the biggest unresolved engineering challenge
Cooling presents perhaps the most difficult problem, since space offers no airflow for dissipating heat the way Earth’s atmosphere does.
Google’s solution relies on thermal interface material connecting the chips to aluminum and copper heat pipes leading outward.
Those components funnel heat toward radiators capable of projecting it directly into the surrounding vacuum of space.
Because of this limitation, the TPUs will only run continuously for about fifteen minutes before requiring a shutdown period.
After each fifteen-minute session, the chips must pause completely while the radiator system catches up on cooling.
Future satellite designs will carry dozens of TPU chips each, communicating across the constellation using precise high-speed laser connections.
Google plans to test two satellites communicating this way sometime in 2027, alongside its previously planned custom satellite launches.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
Google is testing whether AI hardware can survive repeated exposure to space The experimental satellite carries four custom Tensor Processing Units into orbit Radiation testing showed the TPUs could withstand demanding orbital exposure levels Google will launch its first experimental orbital data center satellite on October 1 2026, marking an…
