An upcoming satellite mission will put Google’s custom AI chips into orbit for the first time, giving the company real-world data on whether the hardware can operate reliably under the radiation, temperature swings and other stresses of space.
The flight is the first orbital test for Project Suncatcher, a research effort exploring whether large-scale AI computing infrastructure could eventually be built in space. Google says the prototype satellite will launch next week aboard SpaceX’s Transporter-18 rideshare mission, with satellite company Planet supporting the mission.
Project Suncatcher was announced in November 2025, but the long-term idea remains far from an operational space-based data center. This first mission is focused on a more basic question: whether Google Tensor Processing Units, or TPUs, can survive launch and continue running AI workloads once they reach orbit.
The first test focuses on surviving space
Google has already subjected its hardware to ground tests designed to reproduce some of the conditions it will encounter.
The satellite was shaken along three axes to simulate launch vibrations, while the TPUs were separately exposed to proton radiation at UC Davis’s Crocker Nuclear Laboratory while running AI workloads. Google says its Trillium TPUs withstood a total ionizing radiation dose greater than what the company expects them to receive during a five-year mission.
The orbital test will provide data that laboratory testing cannot fully replicate. Google plans to study how the chips respond to launch stresses, radiation and thermal conditions while operating in space.
Cooling is one of the harder engineering problems. Conventional data centers can move large amounts of air or liquid to remove heat from processors, but a spacecraft operates in a vacuum. Google is testing a system based on heat pipes and radiators to move heat away from the TPUs and release it into space. The design has already been evaluated in a thermal vacuum chamber, with the upcoming flight providing its first test in orbit.
A larger computing network would require satellites to work together
Project Suncatcher ultimately envisions groups of satellites carrying multiple TPUs and sharing AI workloads. Google has pointed to the near-continuous solar energy available in low Earth orbit as one possible advantage of putting compute infrastructure above Earth.
Making that practical would require far more than getting individual processors to work in space. Future satellites would need high-bandwidth connections between one another, along with precise control over their relative positions.
Google plans to test that part of the system next year with two satellites communicating over laser links. The company says those links will need to deliver high bandwidth across relatively short distances while the spacecraft remain in motion, a different problem from many existing space-based optical communications systems.
Planet previously said it would build and operate prototype spacecraft for Project Suncatcher, with the mission intended to test both TPU performance and coordinated operation between satellites.
For now, they aren’t putting a full AI data center into orbit. The upcoming flight is an early hardware experiment meant to determine which parts of that longer-term concept can actually withstand space, with the results feeding into more advanced testing.
Leave a Reply