Google is preparing to send its first orbital data center test to space. A satellite carrying Google's custom AI hardware is scheduled to launch on October 1 as part of the Transporter-18 rideshare aboard a SpaceX Falcon 9 rocket, Ars Technica reported on September 24, 2026. The spacecraft, called MVP, is roughly the size of a refrigerator and holds four of Google's custom TPU AI accelerators, the chips used to train AI models and run inference for AI workloads. Google says ground testing has already been done, but the launch will give the team a fuller picture of the challenges involved.
The satellite's solar panels supply only about one kilowatt of power, enough to run a microwave or a hair dryer, according to The New York Times. Google did not build the spacecraft itself. The AI hardware is Google's, but the satellite comes from Planet Labs, a satellite imagery firm. Google had planned to launch two custom satellites in 2027, but it chose to move faster by integrating its chips into a satellite Planet Labs had already built for an early test.
On the ground, a data center runs thousands of these chips and consumes enormous amounts of power, which is one reason solar-powered AI hardware in orbit is seen as attractive. The lone satellite will test several technologies and operate for only a few months. Google eventually wants networks of orbiting satellites that communicate over high-speed laser links, which would probably require expensive dedicated launches. The 2027 launches remain on the agenda, but the team expects it will be years before Suncatcher moves from project to product.
The test will examine how Tensor chips perform in space. Missions typically use hardware built for extreme temperatures or radiation, but these are the same TPUs Google puts in ground servers. Radiation can damage chips or interfere with calculations by flipping bits, and the vibration and high g-forces of launch create additional risk. Google notes that off-the-shelf hardware has held up in past missions, such as the Ingenuity Mars helicopter.
Cooling the chips may be the largest problem, a point many observers have raised as interest in orbital data centers has grown. Space radiator systems are designed to remove relatively small amounts of heat, while AI accelerators generate far more. Google's design connects the chips to aluminum and copper heat pipes through a layer of malleable thermal interface material, conducting heat into a radiator that releases it into space.
Google will run Gemini models on the TPUs during the test, but not continuously. The cooling system can operate in bursts of about 15 minutes, after which the TPUs must shut down so the radiators can catch up. Google hopes the mission will show what works and what does not, helping it refine designs for later flights.
More hardware news from TechManNews.
