Google to Launch First AI-in-Space Satellite on SpaceX Transporter-18 Rideshare

Google will launch a prototype satellite carrying its AI chips into orbit next week aboard a SpaceX Transporter-18 rideshare mission, marking the first in-space test of its Project Suncatcher moonshot to build AI data centers in orbit.

Announced last year, Project Suncatcher is a long-term research effort exploring whether low Earth orbit could support scalable machine learning infrastructure. Satellites in orbit receive near-constant sunlight and can generate up to eight times more solar power than on Earth, leading Google to investigate whether satellite clusters could eventually handle large AI workloads in space. The test satellite was developed in partnership with Planet to answer one key question, “Can our AI hardware operate in space?”

The first challenge is surviving the launch. A trip to low Earth orbit takes about 10 minutes, with sustained acceleration reaching up to 10 times the force of gravity, while individual components such as TPU chips can experience forces of 50 to 100 g. Google said it vibration-tested the satellite across all three axes to replicate launch conditions and was pleasantly surprised that the hardware survived.

Radiation is the next major hurdle. Google ran AI workloads on TPUs inside a proton beam facility at UC Davis’s Crocker Nuclear Laboratory, monitoring for errors such as bit flips caused by solar events and cosmic rays. Initial results showed the Trillium TPUs performed remarkably well, surviving a total ionizing dose greater than what they would experience during a five-year space mission.

Cooling creates another challenge. TPUs produce significant heat in a small area, but there is no airflow in a vacuum, meaning heat can only escape through radiators. Google is combining heat pipes and radiators, has tested the system in a thermal vacuum chamber, and will evaluate how the design performs in orbit before making further improvements.

The long-term goal is much larger. Future Suncatcher satellites would each carry dozens of TPU chips in orbital clusters, connected by lasers capable of delivering very high bandwidth over extremely short distances. Keeping those connections aligned requires extreme precision, similar to hitting a coin-size target from miles away while both points are moving. Google plans to test the laser links with two satellites in orbit in 2027.

This mission also places SpaceX at the center of a second AI-in-orbit race. Musk has been pitching his own orbital data centers as part of SpaceXAI, with the company talking about orbital compute satellites targeted for first launches next year. The company is also stacking up missions, and SpaceX is targeting Monday, September 28 for Starship Flight 14, its first planned orbital attempt.

If Google’s TPUs survive the upcoming mission, the next major milestone will be the 2027 two-satellite test, giving the race to place AI computing infrastructure in orbit its first real hardware data point.