Google’s Suncatcher Aims to Build Orbital AI Data Centers Powered by the Sun

Google has announced a research initiative called Suncatcher, which aims to create orbital data centers for artificial intelligence. The project is based on compact satellite clusters equipped with Tensor Processing Units (TPUs) and interconnected by high-speed optical communication links.

A key advantage of the orbital location is access to virtually unlimited solar energy. The satellites will be placed in a special orbit that ensures almost constant sunlight, allowing them to generate up to 8 times more energy compared to terrestrial data centers.

To operate effectively, the satellites must be positioned extremely close to each other—for instance, a cluster of 81 satellites within a sphere with a radius of about one kilometer, where the distance between them would be only 100-200 meters. This is necessary to enable ultra-fast inter-satellite links, which Google estimates could reach speeds of tens of Tbps. In lab tests on a prototype, engineers have already achieved a speed of 800 Gbps in each direction.

Special attention is being paid to the radiation hardening of the equipment. Tests have shown that the memory in the TPU v6e (Trillium) can withstand radiation doses 2.5 times higher than expected, confirming the feasibility of the concept.

The project’s economic viability directly depends on reducing launch costs. Google predicts that by the mid-2030s, the price of launching cargo into orbit could fall below $200 per kilogram, making orbital data centers competitive. The first step will be a test mission in collaboration with the company Planet, which will see two prototypes sent into orbit in 2027 to validate the technologies. Among the unresolved challenges are efficient heat dissipation, high-speed communication with Earth, and ensuring overall system reliability.