Musk Predicts All Compute Will Move to Space as Google Prepares Orbital AI Test

Elon Musk has predicted that the bulk of the world's computing capacity will eventually shift into outer space, weighing in as Google unveiled plans for an orbital test of AI hardware. Sharing a video featuring Google's Senior Director for Paradigms of Intelligence, Travis Beals, Musk said the share of compute hosted in space would "obviously round up to 100% of all compute."

Musk had earlier posted on X on September 13 that he was "highly confident" SpaceX would launch Nvidia VR NLV72 AI computers into space next year. The effort forms part of SpaceX's Starmind concept, which envisions AI data centres in orbit. Under that vision, satellites carrying localised compute would draw on readily available power and cooling in space and beam data back through high-bandwidth lasers to the Starlink constellation.

Google, meanwhile, announced that its own research moonshot, Project Suncatcher, is set to conduct its first in-orbit test. A prototype satellite carrying Google's Tensor Processing Units (TPUs) is scheduled to lift off aboard a SpaceX Falcon 9 rocket on October 1, with the mission designed to assess how the chips perform in space.

Google describes Project Suncatcher as a long-term exploration of whether space could one day host scalable machine learning infrastructure. In low Earth orbit, satellites can tap near-constant sunlight, generating as much as eight times more solar power than is possible on Earth. Over time, multiple constellations of satellites could be linked to handle larger AI workloads in orbit.

In the video, Beals said the team's first instinct was to look for reasons the idea could not work. "It turns out that it is possible," he said, describing the project's core question as how to scale AI infrastructure efficiently through orbital data centres by using the abundant energy available from the sun.

Beals said the team examined various orbits, including geosynchronous orbit, before settling on the dawn-dusk orbit as optimal because solar panels there receive sunlight almost around the clock. At its smallest scale, he said, a single satellite would carry dozens of TPU chips, and larger problems would be tackled by linking many such satellites into clusters.

Calling the sun the ultimate energy source in the solar system, Beals said that once the focus turns to scaling solar power as simply and efficiently as possible, doing so in space in the right orbit becomes the natural answer. Google has cautioned that treating space as a viable home for scalable AI compute will not happen all at once, and that the first launch is intended to show what works, identify points of failure and feed those lessons into future missions.