## Google Plans Orbital AI Datacenters Powered by Space-Based Solar Energy

Google on Tuesday announced a new ambitious project – launching constellations of solar-powered satellites equipped with its tensor processing units (TPUs) to create orbital AI datacenters.

According to a recent blog post, Google executives believe that “in the future, space might potentially be the best place to scale AI compute.” They explain that solar panels in space can be eight times more efficient and provide continuous power compared to those on Earth.

The availability of energy is a growing limitation for building terrestrial datacenters, making the prospect of abundant, clean, and uninterrupted power incredibly appealing.

however,Google acknowledges notable hurdles to realizing this vision. A major challenge is securing sufficient rocket launches to deploy a ample infrastructure into orbit. While spacex anticipates over 140 launches this year, launch capacity remains limited.

Google Aims for Terabit-Per-Second Satellite Links with Spatial Multiplexing

Google is exploring spatial multiplexing to dramatically increase data transfer rates between satellites. By deploying this technique, which utilizes multiple autonomous data streams, Google anticipates achieving connection speeds of “tens of terabits a second.”

the company has already demonstrated the effectiveness of the technology using 800 Gbps optics, but acknowledges that these systems were energy-intensive.

“Achieving this kind of bandwidth requires received power levels thousands of times higher than typical in conventional, long-range deployments,” Google explained.”As received power scales inversely with the square of the distance, we can overcome this challenge by flying the satellites in a very close formation – kilometers or less.”