Google puts four TPUs in orbit to test its space-compute plan

Why it matters

The prototype puts Google's proposed orbital AI infrastructure through its first in-space hardware test. The results will bear on whether the chips, communications links and economics behind its satellite-compute concept can move beyond research.

Google put four Tensor Processing Units (TPUs) into orbit on October 1st, sending a prototype satellite for Project Suncatcher on SpaceX's Transporter-18 rideshare mission. In a post on X, Google said it had confirmed contact with the satellite and that it was operating as expected. The mission is a hardware test for a research program exploring whether machine-learning infrastructure could eventually operate in space.

https://x.com/Google/status/2105803583648100611

Travis Beals, Google's senior director of Paradigms of Intelligence, is leading the project publicly. Google and Planet built the prototype together; SpaceX's Transporter-18 mission carried it as one of 130 payloads on a dedicated small-satellite rideshare flight from Vandenberg Space Force Base in California. SpaceX's deployment schedule lists the satellite as Project Suncatcher M1, deployed roughly 61 minutes after liftoff.

The immediate experiment is deliberately narrow. Google plans to collect data over the coming weeks on how the TPUs withstand launch stress, radiation and the temperature extremes of orbit. That tests whether high-performance AI chips can work in the space environment; it does not demonstrate a functioning orbital data center, commercial cloud service or customer access. Google's announcement says the prototype is the first step in a long-term research effort.

Google's research paper sketches the eventual concept: a network of solar-powered satellites in close formation, connected by free-space optical links. One illustrative design uses 81 satellites in a cluster with a one-kilometer radius at an average altitude of 650 kilometers. The paper describes that as a proposed configuration, not a deployment plan. Google's 2025 launch announcement had set an initial milestone of sending two prototypes with Planet by early 2027; the first satellite has now flown, while Google has not described the status or timing of a second prototype in its latest announcement.

The research includes testable engineering claims and assumptions about future costs. Google says solar panels in its proposed dawn-dusk, sun-synchronous orbit could receive up to eight times more annual solar energy than panels at Earth's mid-latitudes. The concept would keep satellites in near-continuous sunlight, but a constellation would still need to move compute data between satellites and back to ground stations. The paper says a bench-scale optical-link demonstration transmitted 800 gigabits per second in one direction. That was a ground experiment, not a result from the satellite now in orbit.

Radiation is another early hurdle. In its paper, Google reports testing its V6e Trillium TPU on the ground with a proton beam. The test found no hard failures from total ionizing dose up to the maximum dose applied to a single chip, and memory irregularities began at a dose nearly three times Google's estimate for a shielded five-year mission. Those findings inform the project, but the in-orbit tests will show how the hardware behaves in actual operating conditions. Google's announcement does not specify which TPU generation is aboard the satellite.

The economics are further out. Google's paper estimates launch costs could fall below $200 per kilogram by the mid-2030s, a projection that depends on sustained declines in launch prices. The team says that at that level, launching and operating a space-based data center could approach the reported energy costs of an equivalent terrestrial facility on a per-kilowatt-per-year basis. Google presents this as an estimate, not a current cost comparison or a commercial business case. The paper also identifies thermal management, ground communications and on-orbit reliability as unresolved engineering challenges.

Google announced Project Suncatcher in November 2025 as a research moonshot. Its October 1st update moves one piece of that research from simulation and ground testing into orbit. The next evidence will be whether four TPUs can keep functioning under real space conditions. Google's larger proposal depends on solving the TPUs' reliability problem alongside satellite networking, heat rejection and launch costs before orbital compute can be compared with the data centers it would be meant to complement.