The next computing frontier may no longer be terrestrial.
California
Google plans to launch its first satellite carrying artificial intelligence chips on October 1, turning Project Suncatcher from a long term research concept into an orbital experiment. The refrigerator sized prototype, called MVP, will travel aboard a SpaceX Falcon 9 from Vandenberg Space Force Base in California. Built with satellite imaging company Planet, the spacecraft carries four Google Tensor Processing Units designed to train and run AI models. Together, they provide computing capacity comparable to a single terrestrial data center server.
The experiment is driven partly by a constraint becoming increasingly important for artificial intelligence: energy. Training and operating advanced AI models requires enormous amounts of electricity, placing growing pressure on power grids and data center infrastructure. Google argues that solar panels in low Earth orbit could generate significantly more energy than equivalent systems on the ground because they avoid atmospheric interference, clouds and conventional day night cycles. Space, under that logic, becomes not simply a communications environment but a potential energy platform for computing.
MVP is not yet an operational orbital data center. Its planned one year mission is primarily intended to determine whether AI processors can survive launch vibration, cosmic radiation and extreme temperature fluctuations. Cooling represents one of the most difficult engineering challenges because conventional air based convection is unavailable in a vacuum. Google has developed radiator panels and conductive materials to move heat away from the chips, but the TPUs are currently expected to operate for only about 15 consecutive minutes before requiring time to cool.
The project could expand substantially if those initial tests succeed. Google plans additional satellites in 2027 to experiment with high speed laser links capable of connecting spacecraft into a distributed computing system. A future architecture could involve more than 80 satellites functioning collectively rather than as isolated processors. Such a system would transform orbital infrastructure from individual machines into a networked computational environment.
Google is not alone in exploring the concept. SpaceX has proposed large constellations of computing satellites connected to artificial intelligence activities, while Blue Origin is pursuing its own orbital infrastructure ambitions. The emerging competition suggests that the energy demands of AI are beginning to influence not only terrestrial data center construction but also the strategic economics of space.
Google does not expect orbital computing to become commercially useful immediately. But if launch costs continue falling and space based energy proves economically viable, the geography of artificial intelligence infrastructure could eventually expand far beyond the surface of the planet.
Información que anticipa futuros. / Information that anticipates futures.