The AI industry is on a $3 trillion, planet-spanning building spree, with “earthbound datacenters” stretching from India to Texas, Lincolnshire to Brazil. In response, Google’s “Project Suncatcher” is offering a radically different vision: “compact constellations” in space.
This contrast between “sprawl” and “compact” is at the heart of Google’s new proposal. The terrestrial model requires massive tracts of land, connection to power grids, and access to vast water resources for cooling. This “sprawl” is the source of “rising concern” over AI’s environmental impact.
Google’s “moonshot” research envisions “compact constellations of solar-powered satellites.” This model replaces land with altitude, grid power with 8x-more-efficient solar panels, and water cooling with thermal radiation in a vacuum. It is a “minimalist” approach to infrastructure.
This “compact” vision, however, has a high cost of entry. Each constellation of 80 satellites must be launched on rockets, which emit “hundreds of tonnes of CO2.” And this “compact” orbital formation is exactly what astronomers fear, as it creates a dense cluster of light-polluting objects.
With prototypes set for 2027, Google is betting that the long-term benefits of a “compact” orbital model will outweigh both the $3 trillion “sprawl” on Earth and the new environmental costs it creates in the sky.