Orbital AI: Google Unveils Project Suncatcher to Launch TPU Data Centers into Space

Google is exploring a space-based twist on AI data centres with its Project Suncatcher, aiming to deploy TPU-equipped satellites to perform AI computations in orbit and harness solar power around the clock. The plan envisions data centres in space that could slash terrestrial cooling and operating costs, with satellites fitted with high‑efficiency solar panels claimed to generate power continuously and, according to Google, solar panels in space are up to eight times more efficient than those on Earth.

However, the project faces major hurdles, including maintaining satellite communication links capable of tens of terabits per second and managing collision risks within a tightly coordinated constellation. Radiation exposure is another obstacle, and Google has already conducted radiation tolerance tests on its Trillium TPUs, confirming they can endure the total ionising dose equivalent to a five‑year mission without permanent damage.

The firm estimates that by the mid‑2030s, energy expenses for space‑based data centres could become comparable to terrestrial facilities, and Google plans to collaborate with Planet Labs to launch prototype satellites by 2027 to evaluate performance. According to Google, the initiative would eventually expand AI infrastructure while exploring a novel, space‑based energy model.