Ground-based AI data centers consume a lot of energy. To avoid this, Google is planning to build a data center in space. The company has announced Project Suncatcher.

Google has made a mega plan to launch an AI data center in space. Despite numerous challenges, including communication and radiation, Google plans to launch two prototype satellites in early 2027. Just recently, the first phase of an underwater AI data center in China was completed.

Why does Google want to build a data center in space?

AI data centers require a lot of energy. This has raised environmental concerns about data centers on land. In light of this, Google wants to utilize the virtually unlimited energy available in space. Solar panels in space can operate continuously with 8 times more productivity than on Earth. A Google official stated that space may be the only place to expand AI computing in the future.

Multiple Satellites to Be Launched Simultaneously

In Project SunCatcher, Google will launch multiple solar-powered satellites simultaneously. These satellites will be equipped with Tensor Processing Units (TPUs), which will be used to create an AI data center in space. However, this will face several challenges. Google is currently testing how to protect TPUs from space radiation for extended periods. Communication will also be a major challenge. To achieve the same efficiency as a ground-based data center, a space center will require very fast networking. To achieve this, Google plans to transfer data between satellites using laser beams.

How Much Will It Cost?

Today, the cost of sending anything into space is enormous, but Google's calculations indicate that rocket launches will decrease by the 2030s. For this reason, running a data center built in space will cost almost the same as running a data center built on the ground.