Google says its Suncatcher prototype — a satellite designed to test whether data centers could eventually run in orbit — is online and communicating with the ground after launching earlier this month alongside more than 100 other spacecraft on a SpaceX Falcon 9 rocket.
The satellite carries four Google Tensor Processing Units, though only one can run at a time; it generates up to 1 kilowatt of electricity and, because of the chip's power demands, can operate for only about 20 minutes at a stretch, a constraint that complicates how often it can communicate with engineers on the ground. After roughly three weeks of subsystem testing, the team's first real workload was almost whimsical: they asked Google DeepMind's Gemma model, preloaded on the satellite, to write a haiku about space. It replied: "Blue marble below / Data streams in silent grace / Starlight feeds my core."
Why anyone would put a data center in orbit
Behind the novelty is a serious engineering question. Space offers nearly constant solar power and no need for cooling infrastructure, but radiation can cause "bit flips" that silently corrupt data without tripping normal monitoring tools. Ground testing suggested the radiation risk would be manageable for simple inference tasks but a bigger problem for training large models; the in-orbit tests are now underway to confirm that. A Google Research analysis published earlier in October estimated that SpaceX would need roughly 1,800 fully reusable Starship flights to bring launch costs down enough to make orbital data centers economically viable — a dependency that ties Google's moonshot directly to the pace of SpaceX's own rocket program, in which Alphabet holds a significant investment stake.
If the current satellite's TPUs hold up under thermal and radiation stress, Google plans to launch two more satellites with more advanced chips in 2027, this time demonstrating high-speed optical links between satellites meant to let multiple spacecraft share a single distributed workload. Google frames Suncatcher as infrastructure for the 2030s, not a near-term product — a useful caveat for a week in which Alphabet's own stock fell sharply after the company reported that its AI investments in search were weighing on ad-revenue growth.
