AI News Feed
Market watch
AI Chips & Compute

Google's First Project Suncatcher AI Satellite Set to Launch Next Week on a SpaceX Rocket

Google will launch a prototype satellite carrying four TPUs on a SpaceX Falcon 9 next week to test whether AI chips can operate in orbit.

The prototype, called MVP and roughly the size of a refrigerator, carries four of Google's custom Tensor Processing Units, the accelerators the company uses on the ground to train AI models and run inference on them. Planet Labs PBC, which normally builds satellites for high-resolution imagery of Earth's surface, developed the spacecraft, while Google supplied the chips. Its solar panels are expected to generate about a kilowatt, roughly the power needed to run a microwave or a hair dryer. The satellite will fly as one part of a larger payload on the Falcon 9, according to SiliconANGLE.

In a blog post, Travis Beals, senior director of Google's Paradigms of Intelligence research team, wrote that the first launch is about seeing what works, identifying points of failure and applying those findings to future missions.

Google announced Project Suncatcher in November 2025, after SpaceX founder Elon Musk outlined his own plans to launch space-based data centers for AI workloads. Google's stated aim is to design orbital AI data centers that could serve as a lower-cost alternative to ground-based facilities, which use large amounts of energy and water and have become increasingly contentious, with many people opposing their construction. Like SpaceX, Google wants to make use of the solar energy available beyond Earth's atmosphere to lower the cost of AI training and inference.

The mission is intended to test several conditions Google considers critical. The chips must survive a roughly 10-minute launch marked by intense vibration and forces up to 10 times stronger than Earth's gravity. They also have to withstand higher levels of radiation from cosmic rays and solar events, which could damage the chips' electronic circuit boards or corrupt data.

According to Beals, Google has already tested its Trillium TPUs at UC Davis's Crocker Nuclear Laboratory, exposing them to a proton beam while they ran AI workloads. Researchers monitored the workloads for errors and found the chips performed "remarkably well," despite receiving a radiation dose greater than what they would be expected to encounter over a five-year space mission.

Cooling remains an unresolved problem. The chips generate substantial heat at full capacity, and the vacuum of space offers no airflow, so engineers cannot rely on traditional cooling systems. Beals said the team is working on several approaches, including a combination of heat pipes and radiators, and has tested the technology in a thermal vacuum chamber that simulates both the thermal and vacuum environment of space. The prototype launching next week will not be able to keep pace with its four TPUs: Beals wrote that the chips will run workloads in bursts of about 15 minutes before they must be shut down to cool off.

Google plans to put two additional satellites into orbit next year to test the high-performance lasers it envisions using for communications. The company has said that a full constellation, if it reaches that stage, would likely require dedicated SpaceX launches.

Editor's Summary

Google is set to launch its first Project Suncatcher prototype on Oct. 1 as part of a SpaceX Falcon 9 payload, carrying four TPUs built to be tested against launch stress, radiation and the vacuum of space. The mission will inform designs for a planned orbital AI data center constellation, which Google positions as a cheaper alternative to energy- and water-intensive ground facilities. Google is pursuing the effort alongside SpaceX, which has announced its own space-based AI infrastructure plans.