Google has set a clear date for its “first milestone” in the AI space race: “early 2027.” This is when the company plans to launch “two prototype satellites,” kicking off the experimental phase of its “Project Suncatcher” initiative.
These prototypes are not the final datacenters; they are testbeds. Their primary mission is to see if Google’s engineers can solve the “significant engineering challenges” that the company itself has acknowledged. This is about data gathering, not service delivery.
The key systems on trial will be “thermal management” (can a TPU be cooled in a vacuum?), “high-bandwidth ground communications” (can the laser links reliably transmit data?), and “on-orbit system reliability” (can the satellite survive for years without repairs?).
This 2027 launch will be a high-stakes moment. It puts Google on a timeline that is significantly behind competitors like Starcloud, which is launching Nvidia chips this month. This suggests Google is taking a more cautious, research-focused approach to building a “scalable” system.
The results from these two prototypes will determine the future of Google’s entire “moonshot” project. If they fail, the concept of orbital AI may be set back by years. If they succeed, they will unlock the path to the 80-satellite constellations and a 2035 target for economic viability.