Google Tackles Energy Demands with Space-Based AI Compute Hardware

Google's ambitious Project Suncatcher blasted off into space on Thursday aboard a SpaceX Falcon9 rocket from Vandenberg Space Force Base in California. The mission marks the beginning of Google's long-term research program to explore the feasibility of scaling AI infrastructure in space.

The first step is to test whether Google's machine learning processors can survive the trip to low earth orbit, where they will be exposed to 'near-constant sunlight' and nearly eight times more solar power than on Earth. The ultimate goal is to build AI compute hardware into satellites and launch them into orbit, potentially reducing data center energy demands.

However, experts warn that launching data centers into space creates new environmental problems and compounds old ones. Dr. Aaron Boley, Professor of Astronomy at the University of British Columbia, notes that every ton put into orbit requires ten to thirty tons of fuel, which can dump a tremendous amount of exhaust into the upper atmosphere.

Google is not alone in its orbital compute ambitions, as SpaceX and Nvidia have partnered to develop Starmind AI1, a proposed fleet of one million satellites. The Indian company TakeMe2Space and micro-gravity life sciences firm Varda Space Industries also participated in the SmallSat RideShare Program with Google.