Google is preparing to send artificial intelligence processors into low Earth orbit as part of Project Suncatcher that was announced in 2025. The research will examine whether machine-learning computing could occur in space and hardware can operate reliably in orbit.
According to a Google blog, the first test is scheduled as part of SpaceX’s Transporter-18 rideshare mission that involves satellite company Planet. The test will place Google Tensor Processing Units (TPUs) in orbit to collect data on how the processors perform under the vibration, radiation and temperature conditions of space. Google has argued that satellites in low Earth orbit could have access to substantially more solar energy than comparable infrastructure on Earth because of long periods exposed to sunlight.
Readers will recall Project Suncatcher was mentioned in the SPECIAL FEATURE | Space, the new data center frontier as an investigation into the potential use of satellite constellations for large-scale AI computing and data centers.
Testing the hardware
A rocket launch exposes spacecraft to intense vibration and acceleration while the spacecraft can experience sustained loads of 10 times the force of gravity during launch, while individual components can experience forces of up to 50 to 100 g. The company subjected the satellite and its components to vibration tests designed to reproduce the conditions of a launch that covered all three axes of movement.
Radiation is another concern since outside the protection of Earth’s atmosphere, electronics are exposed to solar particles and cosmic rays that can cause errors in computer hardware.
Google tested its Trillium TPUs using a proton beam at the University of California, Davis’ Crocker Nuclear Laboratory while the chips were running AI workloads. Researchers monitored the systems for errors, including bit flips, caused by radiation.
The initial tests showed that the TPUs could withstand a total ionizing dose greater than the radiation expected during a five-year space mission. The orbital test will provide data that cannot be reproduced fully in ground-based facilities.
Keeping the chips cool
Thermal management is another engineering problem for computing systems in orbit. TPUs produce significant amounts of heat, but conventional cooling methods that rely on air movement do not work in the vacuum of space. Instead, heat has to be transferred to radiators and released as thermal radiation.
Google has been testing a cooling system using heat pipes and radiators in a thermal-vacuum chamber designed to replicate conditions in space. The upcoming mission will provide a test of how that system performs in orbit.
Connecting satellites
Google’s longer-term concept involves multiple satellites operating as a distributed computing system and future satellites could carry dozens of TPU chips and communicate using laser links.
Those links would need to provide high bandwidth over relatively short distances while the satellites move rapidly around Earth. The satellites would also need accurate information about their own positions and those of neighboring spacecraft to maintain the connections. Google plans to test this part of the project in 2027, when it expects to place two satellites in orbit.
The first mission focuses on whether the computing hardware and its supporting systems can operate reliably in space. Data from the test is expected to guide the design of later missions as Google investigates whether orbital infrastructure could eventually support larger AI workloads.