According to SpaceNews, Planet has announced a partnership with Google on an orbital AI data center project called Suncatcher. The first phase involves launching two specially modified Planet spacecraft, equipped with Google’s tensor processing units (TPUs), by early 2027. These satellites will test TPU performance in space and demonstrate high-bandwidth links between them. Planet CEO Will Marshall called the deal a “competitive win,” leveraging the company’s experience from launching over 600 satellites. He emphasized the project is currently in R&D, aligning with Planet’s own next-gen “Owl” imaging satellite program. The long-term vision, as outlined by Google, involves massive clusters of satellites, potentially numbering in the thousands.
The Space Energy Play
Here’s the core idea that’s got everyone from Google to SpaceX excited: free, nearly continuous solar power. On Earth, data centers are hitting a wall. They need insane amounts of energy for AI compute, and the grid and cooling infrastructure are struggling to keep up. In a dawn-dusk sun-synchronous orbit, a satellite gets almost constant sunlight. No clouds, no night. The power problem, in theory, just vanishes. Google’s own research blog post talks about clusters of 81 satellites flying in tight formation. It’s a compelling vision. But, and this is a huge but, theory is one thing. Actually making it work is a whole other ballgame.
The Not-So-Small Challenges
So what’s the catch? Basically, space is a brutally hostile environment for delicate electronics. The number one issue is heat. On Earth, you cool a server rack with giant chillers and water. In space, your only option is to radiate heat away into the vacuum. Shedding the immense thermal load from a rack of TPUs or GPUs is a monstrous engineering challenge Marshall himself mentioned. Then there’s radiation. Cosmic rays and solar particles can literally flip bits in computer memory, causing errors and crashes. You need serious hardening, which adds cost, weight, and complexity. And let’s not forget the sheer logistics of building, launching, and maintaining thousands of these things. It’s not just plugging in a server and turning it on.
A Crowded Orbit
Planet and Google are far from alone in this dream. Elon Musk has explicitly said SpaceX is looking at orbital data centers. Jeff Bezos predicts they’ll be cheaper than ground-based ones “in the next couple of decades.” Startups like Starcloud and Aetherflux are already testing components, with Aetherflux targeting 2027 for its first “Galactic Brain” nodes. This isn’t just sci-fi speculation anymore; it’s becoming a real R&D race. The companies that solve the thermal and radiation problems first, and drive down launch costs, will own this new frontier. For an industry built on rock-solid reliability, moving to the most unreliable environment imaginable is a fascinating bet. It reminds me of the early cloud computing skeptics—sometimes the biggest problems require the most radical changes in perspective.
Why Planet? Why Now?
Look, Planet isn’t a cloud computing giant. So why are they a key player here? It comes down to infrastructure and timing. Marshall nailed it: there are only a handful of companies with real experience in scaled satellite constellations. Planet has launched and operated hundreds. They have the buses, the launch contracts, the ground stations, and the operational know-how. For Google, partnering with them is faster and cheaper than building that capability from scratch. For Planet, this is a strategic pivot. They’re moving from being just an “imaging company” to a “space and AI company.” This R&D project could open a revenue stream vastly larger than selling satellite photos. If the future of compute is truly heading upstairs, Planet wants to be the landlord. And if you think about the hardware demands for controlling these orbital clusters, it’s a realm where rugged, reliable computing is paramount—the kind of industrial-grade reliability that companies like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, specialize in for extreme terrestrial environments. The principles are similar, even if the setting is light-years apart in difficulty.
The Long Game
Is any of this going to happen soon? No. Everyone involved is clear this is a decade-plus vision. The 2027 demo is just a critical first step to prove the core tech works. But the statements from CEOs like Sundar Pichai and Will Marshall are telling. They’re not treating this as a side project; they’re framing it as an inevitable evolution. As launch costs fall with reuse and satellite buses become more commoditized, the economic equation shifts. The question changes from “Can we do it?” to “Why *wouldn’t* we do it?” I think that’s the real significance of Project Suncatcher. It’s not about two test satellites. It’s about planting a flag for a post-terrestrial cloud. The race to build the first functional orbital data center cluster is officially on.
