Google Launches First Orbital TPU Test as Paper Estimates Starship Launch Demands

The satellite was built by Planet Labs. Once commissioned, it will fire up its TPU in 15-minute bursts to avoid straining the satellite's power and thermal management systems, and run a series of models to see whether anything goes wrong. Travis Beals, the Google executive managing Project Suncatcher, said ground testing is not a complete substitute for flight. "We've done testing on the ground, but you know, there's no test that's completely as good as the real thing," Beals said.

Project Suncatcher is Google's plan to develop large-scale compute clusters in orbit around Earth. The current satellite is based on a standard platform built by Planet Labs, but the two companies are working on a demonstration expected to fly next year involving two satellites more purpose-built for advanced compute. Those future versions will attempt to collaborate through a laser communications link.

The SpaceX rocket carried more than 100 different payloads, including missions from Satlyt and Cowboy Space Company. Google's initiative differs from those startups, and from SpaceX itself, because it is a long-term project. Google envisions a network of 81 satellites flying in close formation and processing in parallel. Beals said bandwidth and latency between TPUs matter greatly for multi-rack workloads, and Google is looking ahead not only to workloads that exist today but to where they will be in five years. He also said the rockets required to scale up orbital data centers in a cost-effective way do not yet exist.

On Thursday, Google also released a peer-reviewed version of its white paper on orbital data centers, which will be published in Joule. The paper stresses that its analysis is not an economic feasibility study, but it offers a picture of how Google sees rockets becoming cheaper over time. Google is looking to SpaceX to get its spacecraft off the ground and is also a major investor in SpaceX.

The authors argue that SpaceX has achieved a price-reducing learning curve of about 20% a year since it launched the Falcon 1 rocket, making it reasonable to expect launch prices close to $200 per kilogram by 2035. Based on the amount of payload launched by Falcon 9, they think a similar cost-reduction trajectory would require Starship to fly 370,000 tons of payload into orbit. That would take about 1,800 launches over the next ten years, or 180 a year, if Starship can fly 200 metric tons on each mission.

That is a large number for a vehicle that has never flown more than five times in a year. SpaceX predicts it will fly far more than that; Elon Musk has suggested Starship could achieve an hourly flight rate in 2029. Google's updated research also suggests its chips will likely survive the radiation of space. The company had to redo tests blasting the chips in a particle accelerator after realizing the configuration of the chips provided more shielding than they would actually experience. This produced slightly more errors in the chip's logic circuitry, but Google remains confident its chips can handle large inference workloads in orbit for the five-year lifespan of a satellite.

"The error rate is very low if you're thinking about typical inference operations, right? Like one in a million," Beals said. "On the other hand, it was already problematic for doing, say, some mega-scale training run where you're going to have many thousands of chips running for months."