Project Suncatcher’s first prototype was reported to have reached orbit aboard SpaceX’s Transporter-18 on Oct. 1, 2026. Developed by Google with Planet, the spacecraft is a research test of how machine-learning hardware responds to conditions in space—not an orbital data center operating at scale. The planned flight was the subject of earlier pre-launch coverage.
The prototype’s orbital mission is a hardware test
Google says the mission is intended to collect data on how its Tensor Processing Units (TPUs)—chips designed for machine-learning workloads—respond to physical stress, radiation and thermal extremes. The prototype reportedly carries four TPUs and runs a version of Google’s Gemma model for about 15 minutes at a time because of heat-management constraints.
The prototype is reported to use batteries in a standard orbit. A sun-synchronous orbit belongs to Google’s future design concept. The distinction matters: the current flight tests hardware, while the proposed system would explore a different orbital arrangement.
What Project Suncatcher is testing in space
Heat is a central engineering problem. A vacuum has no airflow to carry heat away, so Google describes using heat pipes and radiators to move and release it. The company identifies thermal management and reliability as continuing research challenges; repairs and equipment replacement in orbit also add operational complexity.
Separate ground tests examine how the hardware responds to radiation. Google reports testing V6e Trillium TPUs with a 67 MeV proton beam. In listed tests on a single chip, exposures reached 10 krad(Si) without hard failures attributed to total ionizing dose; HBM-specific stress tests showed irregularities after a cumulative dose of 2 krad(Si).
Google’s explainer uses laboratory footage and animations to describe launch-force and radiation testing. It offers context for the engineering work behind the project.
What Google’s launch-cost model projects
Google’s research paper models a possible decline in the cost of launching payloads to low Earth orbit: prices could reach about $200 per kilogram by the mid-2030s if the model’s assumptions hold. Its learning curve assumes an approximately 20% reduction in price per kilogram each time the cumulative mass launched doubles.
The paper says sustaining that curve would require roughly 180 Starship launches per year. That is a modeled annual cadence, not a confirmed launch schedule. The estimate concerns a possible future launch-cost path; it does not establish that orbital data centers are already economically viable. The prototype’s reported trip, meanwhile, was aboard Transporter-18.
Google plans an optical-link test for 2027
Google plans to put two satellites in orbit in 2027 to test optical links—communications that use light to carry data between spacecraft. The test would address one of the project’s remaining engineering questions: how satellites in a future system could exchange data at high bandwidth.
