Meet Project Suncatcher, Google’s Plan to Put AI Data Centers in Space

Why It Matters

If successful, space‑based AI compute could dramatically lower energy and real‑estate costs for the rapidly expanding AI industry while creating a new market for low‑cost launch services, but it also raises significant technical, regulatory and astronomical‑impact challenges.

Summary

Google unveiled Project Suncatcher, a moonshot to launch solar‑powered satellites equipped with Cloud TPUs that form a distributed AI compute network in low‑Earth, sun‑synchronous orbit. The design relies on free‑space optical links that have demonstrated 1.6 Tbps bidirectional speeds on the ground and aims to keep satellites within a kilometer of each other for high‑speed data exchange. Google is radiation‑testing its v6e TPUs, which survived up to 2 krad, and plans to fly a pair of prototype AI orbiters by early 2027, with cost‑effective operation expected in the mid‑2030s as launch prices fall to around $200 per kilogram. The effort seeks to bypass the energy, land and community constraints of terrestrial data centers by moving compute to space.