Google Is Taking AI to Space. The First Test Is Almost Here

What if your next data center wasn’t located on planet Earth? Google is planning to seriously consider that scenario very soon. Within the framework of Project Suncatcher, the company is set to launch a prototype satellite into space aboard SpaceX’s Falcon 9 on October 1. This satellite, manufactured by Planet, will be equipped with four Google Tensor Processing Units, or TPUs.

At the moment, this is not going to be an operational data center. Google needs to know how its AI hardware performs in space prior to undertaking any more ambitious projects.

The idea has a simple appeal. Satellites in the right orbit can receive sunlight for much longer periods than solar panels on Earth. Google believes this could provide a powerful energy source for future AI infrastructure.

Space Gives Google Power, But It Takes Away the Easy Solutions

But getting the chips up there is just the first step. Google will have to study how the TPUs perform under radiation, during the violent process of launching, and in extreme temperatures. It could be particularly challenging to cool them down. On Earth, air and water are used for cooling down data centers where there are very powerful chips. In outer space, there is no such possibility. The company has already tried out heat pipes and radiators.

Google plans to link together many satellites with an optical connection and build bigger computer clusters in outer space. This was what the preliminary studies were about.

What Could Orbital Data Centers Mean for You?

This is one experiment which we believe is worth considering even though space data centers seem like something of a distant future at this point.

If Google succeeds in addressing all problems related to energy sources and cooling, the infrastructure which is required to be used in AI operations might no longer need land-based locations. It will also help bring energy-hungry computing operations closer to satellite data collection locations.

All of this can become your reality sooner or later but before that, Google needs to show that its chips can survive the journey.

Are you curious about the future of AI infrastructure development? Do not forget to share the article and subscribe to our social media pages. There are many other articles related to data centers, AI, and infrastructures available there.