On 1 October 2026, Google launched a prototype satellite that carries its TPU chips into space. It is the first step in exploring whether space could one day host large-scale infrastructure for machine learning.
- Google launched a prototype satellite carrying its TPU chips on 1 October 2026, built with Planet and flown on SpaceX's Transporter-18 mission.
- The idea is to power AI computing with sunlight in orbit, where Google says satellites can generate up to eight times more solar power than on Earth.
- In Google's lab tests, its Trillium TPUs survived more radiation than a five-year mission in orbit would give them.
- This is an early test: over the coming weeks, Google will check how the chips cope with launch stress, radiation and extreme temperatures. Space data centres are still far away.
AI compute consumes huge amounts of electricity and water, because it needs GPUs and TPUs in large numbers to answer users quickly. Google has taken a first step towards addressing this problem: in partnership with Planet Labs, it launched a prototype satellite on SpaceX's Transporter-18 rideshare mission. The idea is to supply computing power from sunlight. The project is called Project Suncatcher, and Google calls it "a long-term research moonshot".
As Google wrote in its blog post:
"Today, our prototype satellite for Project Suncatcher, built in partnership with Planet, launched into orbit aboard the Transporter-18 rideshare mission with SpaceX. Our team has confirmed contact with the satellite and it is operating as expected.
This is the first step in a long-term research moonshot exploring whether space could one day host scalable machine learning infrastructure. Over the coming weeks, we'll gather in-orbit data on how our TPUs handle the physical stress of spaceflight and the radiation and thermal extremes of space."
The mission tests how the chips cope with the physical stress of the launch, and how extreme heat, cold and radiation affect the TPUs. Google has published more details in a peer-reviewed paper in the journal Joule.
What does Google's paper propose?
The paper starts from a simple point: the Sun is by far the largest source of energy in our solar system, so it makes sense to ask how future AI infrastructure could use that power directly. Google describes a computing system made of fleets of satellites, each with solar panels, Google's TPU chips, and laser links that carry data between the satellites (free-space optics).
Radiation is one of the biggest risks for chips in space. According to the paper, Google's Trillium TPUs were tested for radiation and survived a total dose equivalent to a five-year mission without permanent failures.
Cost is the other big question. Google's analysis suggests that the price of launching to low Earth orbit may fall to about $200 per kilogram or less by the mid-2030s. If that happens, the cost of the launch, spread over a satellite's lifetime, could be roughly comparable to what a data centre on Earth pays for energy, per kilowatt.
Google admits that many significant challenges remain before this "moonshot" can work. But it argues that, in the long run, it may be the most scalable way to provide AI computing, and that it would put less pressure on land and water on Earth.