For years, "the cloud" was a marketing euphemism for a windowless warehouse in Virginia. Google apparently took the metaphor personally. As of this week, a small slice of its AI hardware is circling the planet at about 650 kilometers up, which is the most aggressive "work from anywhere" policy we've seen yet.
Four TPUs, One Falcon 9, Zero Wi-Fi
On October 1, Google and Planet launched a prototype satellite carrying four Tensor Processing Units on SpaceX's Transporter-18 rideshare mission, which lifted off from Vandenberg Space Force Base aboard a Falcon 9. It's the first step of Project Suncatcher, Google's research moonshot to test whether scalable machine learning infrastructure could one day live in space.
Google says it has made contact with the satellite and it's operating as expected. The craft sits in a dawn-dusk sun-synchronous orbit, and over the coming weeks the team will collect data on how the TPUs cope with launch stresses, radiation, and the thermal extremes of orbit. A peer-reviewed paper on the research has been published in the journal Joule. "Some things can only be tested in space," Google wrote.
Why Anyone Would Do This
The pitch is energy. AI's appetite for electricity is colossal, and a dawn-dusk orbit keeps a satellite in near-constant sunlight, turning solar panels into a much more reliable power source than they are on the ground. Meanwhile, terrestrial data centers keep running into grid limits and local opposition.
The part most hype skips: getting rid of heat in a vacuum is genuinely hard, radiation is not kind to chips, and you can't send a technician with a replacement board. This prototype is a science experiment, not a product launch — nobody's running your chatbot from low Earth orbit next quarter. But the questions it answers will shape whether orbital compute is a real option or an expensive Bond-villain aesthetic.
Either way, it's a delightful reminder that "scaling AI" now includes rocket schedules. The future is weird, and we're here for it.
Your AI workloads may stay firmly on the ground for now, but if you're weighing where and how to run AI for your business, we can help you pick a setup that makes sense today.
Source: Google Blog