The voracious appetite for artificial intelligence is pushing tech giants to seek radical solutions, and Google is now looking to the stars. The company is developing plans to establish datacenters in space to meet the rising demand for AI, a move that could reshape digital infrastructure.
This initiative, known as Project Suncatcher, envisions a future where AI processing occurs hundreds of miles above Earth. The plan involves creating constellations of solar-powered satellites carrying Google’s custom AI processors (TPUs). This network would be interconnected using free-space optical links, essentially using laser beams to transmit massive amounts of data.
This celestial approach is becoming feasible due to the plummeting cost of rocket launches. Google research suggests that by the mid-2030s, the operational costs of an orbital datacentre could be on par with a terrestrial one. This economic shift, combined with the environmental strain from the $3 trillion projected to be spent on earthbound datacenters, makes space an attractive alternative.
One of the most significant advantages is power. In orbit, solar panels can harvest energy far more efficiently than on Earth, providing a near-constant, low-cost renewable energy source. This would also eliminate the need for the vast quantities of land and water used to cool datacenters on the ground, addressing growing environmental concerns.
However, the plan is not without its challenges. Google notes that significant engineering problems related to thermal management and system reliability must be solved. Furthermore, the environmental impact of rocket launches, which release hundreds of tonnes of CO2, and objections from astronomers worried about satellite “light pollution” present serious obstacles.