Google has unveiled the technological blueprint for its “moonshot” plan to put AI datacenters in space. The initiative, Project Suncatcher, relies on a combination of advanced satellite technology, custom processors, and high-speed laser communications to make orbital AI a reality.
The core of the plan involves constellations of around 80 satellites orbiting 400 miles high. Each satellite will be a compact, solar-powered datacentre equipped with Google’s own TPUs (Tensor Processing Units). These chips are specifically optimized for the heavy workloads of training and running AI models.
To power this network, Project Suncatcher will exploit the unique advantages of space. In orbit, solar panels are free from atmospheric interference and weather, allowing them to be up to eight times more productive than on Earth. This provides a constant, powerful, and renewable energy source for the processing-intensive AI tasks.
Connecting these floating datacenters and sending data back to Earth presents a unique challenge. Google plans to use free-space optical links, which transmit information via light or laser beams. This technology allows for high-bandwidth data transmission without the need for physical cables, beaming results back to ground stations.
While the first prototypes are not expected to launch until early 2027, the plan signals a major new direction for data infrastructure. However, Google admits that significant technical hurdles remain, including managing processor heat in a vacuum (thermal management) and ensuring the long-term reliability of complex systems in orbit.