Google is preparing to put its artificial intelligence hardware where no AI chips have gone before: into orbit. On October 1, the company's Project Suncatcher will launch a prototype satellite carrying four of Google's Tensor Processing Units into low Earth orbit, in a first test of whether AI computing can survive — and eventually thrive — in space.
The mission
The small satellite will fly aboard SpaceX's Transporter-18 rideshare mission in partnership with satellite-imaging company Planet Labs, which announced the collaboration on social media this week. Its purpose is purely experimental: gather engineering data on how Google's AI chips withstand the violence of launch, the radiation of space and the extreme thermal swings of orbit.
Why space?
The pitch is power. AI data centers devour enormous amounts of electricity, while satellites in the right orbit can soak up near-constant sunlight — Google says up to eight times more solar power than comparable panels on Earth. Instead of building ever more power-hungry facilities on the ground, engineers could one day bring the computers closer to the Sun. SpaceX and a startup called Starcloud are pursuing similar orbital data-center concepts.
The torture tests
A rocket launch subjects hardware to forces of 50 to 100 times gravity, so engineers put the satellite through intense vibration testing. Radiation is the other enemy: Google blasted its Trillium TPUs with a proton beam at the University of California, Davis's Crocker Nuclear Laboratory, and says early results showed the chips could tolerate a greater radiation dose than a five-year space mission would deliver. Cooling is trickier — there is no air in a vacuum — so the design uses heat pipes and radiators to move heat away from the chips. The company found that simply restarting the processors usually clears radiation-induced errors.
What it isn't
This is not a working orbital data center. It carries four chips — roughly a single server's worth of compute — and its job is measurement, not production. Google says future work includes testing high-bandwidth laser links between satellites in 2027. The real question is not whether chips can survive in orbit, but whether thousands of them can operate there together, economically and reliably.
Sources: BitcoinVersus, Primetimer
The mission
The small satellite will fly aboard SpaceX's Transporter-18 rideshare mission in partnership with satellite-imaging company Planet Labs, which announced the collaboration on social media this week. Its purpose is purely experimental: gather engineering data on how Google's AI chips withstand the violence of launch, the radiation of space and the extreme thermal swings of orbit.
Why space?
The pitch is power. AI data centers devour enormous amounts of electricity, while satellites in the right orbit can soak up near-constant sunlight — Google says up to eight times more solar power than comparable panels on Earth. Instead of building ever more power-hungry facilities on the ground, engineers could one day bring the computers closer to the Sun. SpaceX and a startup called Starcloud are pursuing similar orbital data-center concepts.
The torture tests
A rocket launch subjects hardware to forces of 50 to 100 times gravity, so engineers put the satellite through intense vibration testing. Radiation is the other enemy: Google blasted its Trillium TPUs with a proton beam at the University of California, Davis's Crocker Nuclear Laboratory, and says early results showed the chips could tolerate a greater radiation dose than a five-year space mission would deliver. Cooling is trickier — there is no air in a vacuum — so the design uses heat pipes and radiators to move heat away from the chips. The company found that simply restarting the processors usually clears radiation-induced errors.
What it isn't
This is not a working orbital data center. It carries four chips — roughly a single server's worth of compute — and its job is measurement, not production. Google says future work includes testing high-bandwidth laser links between satellites in 2027. The real question is not whether chips can survive in orbit, but whether thousands of them can operate there together, economically and reliably.
Sources: BitcoinVersus, Primetimer
