Google Tests Its AI Chip in Orbit in First Suncatcher Flight
Google launched its first orbital computing test on October 1, 2026, aboard a SpaceX rocket from California. TechCrunch
The flight was the first time Google flew one of its advanced chips in space. The spacecraft was built by Planet Labs and will test whether a Google Tensor Processing Unit, or TPU, a chip built for AI work, can operate in orbit. The effort is part of Project Suncatcher, Google's plan for large computing groups in orbit around Earth, led by executive Travis Beals.
Survival comes first. Stable operation matters more than speed at this stage.
After early checks, the satellite will turn on its TPU in 15-minute bursts to avoid stress on power and cooling. The short work periods keep power use and heat within the limits of a small satellite frame that was not built to work as an AI server.
The launch was a rideshare. The SpaceX vehicle carried more than 100 payloads, including missions from Satlyt and Cowboy Space Company. Google's prototype flew on a Falcon 9, according to launch reporting. CNBC
This October flight was the first in-orbit test of Project Suncatcher.
A two-satellite demo and an 81-satellite design
Google and Planet Labs are already working on the next step. A demo planned to fly next year will use two satellites built from the start for heavier computing.
Those two spacecraft will try to work together through a laser link. Laser links between satellites can carry a lot of data with low weight and without radio licenses, but keeping the beams locked between two moving satellites is hard. The demo should give early data on whether computers on separate satellites can stay in sync.
The longer-term design described by Google calls for 81 satellites flying in close formation and processing in parallel. The current prototype is modest by comparison. It has enough computing power to answer simple AI questions from space, a capability described before launch. The New York Times
Google has described Project Suncatcher in company materials as an early test to scale AI computing in space and as its moonshot to put AI in space.
Launch cost as the gating factor
Alongside the hardware, Google released a peer-reviewed version of its white paper on orbital data centers for publication in Joule.
The paper centers on launch economics. Google researchers estimate SpaceX has cut launch prices by about 20 percent per year since the Falcon 1. On that path, they expect prices near $200 per kilogram by 2035.
The paper applies the same logic to Starship. The researchers estimate Starship would need to lift 370,000 tons to orbit to follow a cost path like Falcon 9. At 200 metric tons per flight, that is about 1,800 Starship launches over the next 10 years, or 180 per year.
Google is a major investor in SpaceX.
The broader context here is that Google treats launch rate, not chip speed, as the pacing item for orbital data centers. The TPU test can check radiation tolerance, heat behavior and power cycling in orbit. It cannot by itself change dollars per kilogram. Only flight rate and reusability do that.
In my view, the 15-minute bursts and the move to purpose-built satellites point to where the engineering pressure sits. In vacuum there is no air to carry heat away and no grid to absorb power spikes. On the ground, power and cooling shape data center design as much as chips do. In orbit, that tie is tighter.
Worth flagging for software work, the 81-satellite formation and the laser test raise clear questions. Parallel AI work across a formation needs a steady link, a way to isolate faults when a satellite drops out, and software that treats orbital motion as part of the network. The two-satellite test next year should show whether Google and Planet Labs can hold a link steady enough to make that practical.
Looking ahead, the hopeful case is that plentiful solar power and cheaper launch could open a new layer for AI infrastructure. That depends on Starship flying often and cheaply, on spacecraft that handle steady heat, and on AI tasks that fit widely spread nodes working in bursts. No single prototype settles those questions. This launch starts collecting the flight data needed to answer them.
