AI Data Centres in Space? Musk Backs the Idea as Google Sets October 1 Test

Google plans to launch a prototype satellite on October 1 to test how its artificial intelligence chips perform in space, taking an early step towards a prospect that has drawn an expansive prediction from Elon Musk: almost all computing could eventually move beyond Earth.

The mission, scheduled to launch aboard a SpaceX Falcon 9 rocket, is part of Google’s Project Suncatcher, a research initiative exploring whether orbiting satellites could support large-scale AI infrastructure.

Sharing a video featuring Travis Beals, Google’s senior director for Paradigms of Intelligence, Musk wrote: “The amount of compute in space will obviously round up to 100% of all compute.”

That remains a prediction. Google describes its project as a long-term research effort, with the first mission intended to test equipment, identify failures and guide subsequent experiments.

What Will Google Test in Orbit?

The prototype satellite will evaluate how Google’s Tensor Processing Units, or TPUs, perform in space. These are specialised chips used for machine-learning workloads.

The broader proposal involves satellites carrying multiple chips and communicating with one another to handle larger computing tasks. Beals said an individual satellite could carry dozens of TPUs, with groups of satellites eventually linked together in flying clusters.

The October launch is an initial experiment towards that ambition, rather than the deployment of a fully functioning orbital data-centre network.

Google said the mission would focus on establishing what works, identifying points of failure and using those findings to improve future missions.

Why Put AI Computing in Space?

The attraction is access to solar energy.

According to Google, satellites in suitable low-Earth orbits could receive near-continuous sunlight and generate up to eight times more solar power than on Earth.

Beals said the team had considered several orbital configurations before identifying a dawn-dusk orbit as the best fit for its plans. Such an arrangement, he explained, would give the satellite’s solar panels sunlight almost around the clock.

“At the heart of it, Project Suncatcher is a simple idea. How can we efficiently scale AI infrastructure through orbital data centres by leveraging the abundant energy that we can get from the sun?” he said.

The proposal would use that energy to power AI chips aboard satellites, with interconnected clusters allowing the system to take on larger workloads.

Musk’s Bigger Bet on Space Computing

Musk’s response to Google’s announcement follows an earlier claim about SpaceX’s own plans.

In a September 13 post on X, he said he was “highly confident” that SpaceX would launch Nvidia AI computers into space next year.

SpaceX’s Starmind concept envisages orbital AI data centres with computing equipment aboard satellites. According to the company’s description, the satellites would use power and cooling opportunities in space and transmit data through high-bandwidth laser links to the Starlink constellation.

Google’s prototype is also scheduled to use a SpaceX rocket, although the two companies have described their computing initiatives separately.

A Test of an Ambitious Idea

Beals said Google’s initial response to the concept was to investigate why it might fail.

“This idea, that’s crazy, there must be a reason it won’t work. And so the first thing we tried to do is to find that reason,” he said. “It turns out that it is possible.”

The company’s stated approach remains experimental. The October 1 mission is intended to provide evidence about operating its chips in orbit. Building a scalable network capable of handling substantial AI workloads would require further development and testing.

For now, the immediate question is how Google’s hardware performs once it leaves Earth. Musk’s prediction concerns how far that journey could eventually go.

With inputs from ANI