According to Ars Technica, Google is developing Project Suncatcher, a plan to put AI data centers in space using orbiting satellites with specialized TPU processors. The company aims to launch its first prototype satellites by early 2027 and believes space-based computing could become economically viable by the mid-2030s when launch costs potentially drop to $200 per kilogram. Google’s research shows solar panels in orbit could be up to eight times more efficient than Earth-based systems, solving power constraints. The satellites would operate in low-earth orbit with free-space optical links maintaining communication speeds up to 1.6 terabits per second. Testing reveals Google’s latest Trillium TPUs can handle nearly 2,000 rads of radiation, well above the 750 rad requirement for five-year operation.
Sponsored content — provided for informational and promotional purposes.
<h2 id="why-space“>Why even consider this madness?
Look, terrestrial data centers are becoming a massive problem. They’re energy hogs, water consumers, and nobody wants them in their backyard. Google’s basically saying: what if we just… move the problem off-planet? The solar efficiency argument is compelling – eight times more power generation in space is nothing to sneeze at. And let’s be honest, the sheer novelty factor alone probably got this project greenlit.
The physics don’t care about your ambitions
Here’s the thing about space: it’s actively trying to kill your electronics. Radiation, extreme temperature swings, and the vacuum of space make terrestrial hardware cry. Google’s testing suggests their TPUs can handle more radiation than expected, which is promising. But surviving five years in orbit is very different from surviving five years while processing AI workloads 24/7. And those optical links? Maintaining terabit speeds between satellites moving at thousands of miles per hour while staying within a kilometer of each other? That’s some serious precision flying.
The money question
Google’s banking heavily on launch costs dropping to $200 per kilogram by the 2030s. That’s a big assumption. SpaceX and others are driving costs down, but we’re talking about launching not just satellites, but entire data centers. The initial 2027 prototypes will be expensive as hell, and there’s no guarantee the economics will work out. Remember Google’s other moonshots? Some succeeded, many quietly disappeared. This feels like one that could easily join the graveyard if the numbers don’t add up.
Is this actually solving anything?
Part of me wonders if this is just tech industry excess disguised as innovation. Sure, it solves the “not in my backyard” problem for data centers, but it creates new ones. Space debris, astronomical interference, and the environmental cost of frequent rocket launches. And let’s not forget latency – even low-earth orbit adds milliseconds that matter for some AI applications. Google’s own research paper acknowledges many challenges, and their technical document shows they’re still in early testing phases. I’m skeptical, but honestly? If anyone has the resources to try this crazy idea, it’s Google.
