Google is talking to SpaceX about future launches for Project Suncatcher, its plan to put solar-powered AI data centres in orbit, according to Reuters. The pitch is straightforward enough: move compute off Earth, keep it powered by sunlight, and use Tensor Processing Units aboard satellites to build a space-based AI cloud. The harder part is everything else, which is why Google is also lining up Planet Labs for an initial prototype launch around 2027.
Project Suncatcher aims at space-based AI cloud computing
Project Suncatcher is not some moonshot sketch on a whiteboard; it is Google’s research effort to network satellites carrying its own chips into an orbital data platform. That puts it in the same rarefied club as the kinds of ideas that sound absurd until companies with enormous budgets decide they are unavoidable.
The appeal is obvious. Ground-based data centres are expensive to power and cool, while AI demand keeps climbing. Google’s answer is to make the sun the utility bill. The catch is that space infrastructure is brutally capital intensive, and any orbital computing system has to survive launch risk, radiation, and the small matter of getting enough bandwidth to make the whole thing useful.
Why SpaceX matters to Google’s plan
SpaceX is the obvious partner because it controls the launch stack, and launch access is the gating item for any orbital compute project. Reuters said the companies have been in discussions alongside other potential launch providers, while the Wall Street Journal first reported the talks on Tuesday. If Google gets serious about Suncatcher at scale, it will need cheap, frequent launches more than it needs PowerPoint optimism.
- Project Suncatcher: Google’s research effort for orbital AI computing
- Initial prototype launch: around 2027 with Planet Labs
- Core hardware: Google Tensor Processing Units on solar-powered satellites
The timing is awkwardly interesting. SpaceX is heading toward a widely anticipated initial public offering, and orbital data centres are said to be one of the drivers behind those plans. That makes Google both a potential customer and a strategic validation point for a business that has to prove it can do more than launch rockets and talk about Mars.
Musk, Google and the new space AI race
The deal chatter would also be a neat little sequel in Silicon Valley’s ongoing rivalry theatre. Musk helped launch OpenAI in 2015 partly as a counterweight to Google’s AI ambitions, after disagreeing with Larry Page over AI safety. Now Google and SpaceX are circling the same prize: the first credible AI data centre in orbit.
That race is still early, but the direction is clear. Last week, Anthropic said it would use the full computing power of SpaceX’s Colossus 1 facility in Memphis and showed interest in working with the company on multiple gigawatts of space-based orbital data centres. If that sounds wildly ambitious, that is because it is; if it also sounds like the start of a real market, that is because the cloud giants are already acting like it is.
The obvious question now is whether Suncatcher stays a research project, becomes a launch business, or turns into one more expensive proof that AI wants to go everywhere at once. My bet: Google keeps testing, SpaceX keeps listening, and the first real milestone is not a grand orbital fleet but one prototype that proves the economics are even remotely sane.