Google’s “Moonshot” Mentality: ‘Working Backward’ from a Space-AI Future

Google is applying its “moonshot” research philosophy to the AI boom, “working backward” from a future where AI is no longer constrained by Earth. This new initiative, Project Suncatcher, is the direct result of this long-term thinking, positing that space may be the only place to sustainably scale AI.
The company’s vision, described in new research, is to bypass the $3 trillion, resource-heavy terrestrial datacentre boom. Instead of consuming land and water, Google envisions “compact constellations” of solar-powered satellites 400 miles up, running on its own powerful TPU processors.
This “working backward” approach starts with a major problem: AI’s environmental footprint. Google’s solution is to tap into the 8-times-more-productive solar energy available in orbit. This “unlimited” power source is the key to running an “at-scale” AI infrastructure.
This plan, however, is the definition of a “moonshot.” Google’s own “cautionary note” highlights the “significant engineering challenges” that remain. These include thermal management, data transmission via optical links, and ensuring long-term reliability in an environment where repairs are impossible.
With its first prototypes slated for 2027, Google is just beginning this journey. This “moonshot” is a high-stakes bet that the engineering problems of space are more solvable than the resource and energy problems of Earth.