Google and SpaceX are in early talks about an idea that sounds like sci-fi but is edging toward the spreadsheet phase: shifting pieces of Google’s Project Suncatcher into low Earth orbit. Instead of stuffing AI processors into massive server farms on the ground, the concept would park Tensor Processing Units on solar-powered satellites, using nearly constant sunlight for power and the cold of space as a heat sink. The effort has moved from blue-sky research to active conversations with launch providers this week, a shift that could ripple through Bay Area engineering jobs, supply chains, and the ongoing scramble to lock in new ground-based data center sites.
Talks With Launch Providers
As reported by TechCrunch, which cites a Wall Street Journal scoop, Google has been discussing rocket launches with SpaceX and other launch companies. CBS News carried a short summary as the story rippled through the press. So far, neither company has announced a contract, and spokespeople did not immediately respond to requests for comment.
What Project Suncatcher Would Look Like
Google has published research that frames Project Suncatcher as a "moonshot" to network solar-powered satellites that carry its Trillium TPUs and communicate via free-space optical links, according to Google. The company’s preprint paper illustrates a tightly clustered formation of 81 satellites and describes a "learning mission" with Planet Labs, which is expected to fly two prototype satellites by early 2027; the Google preprint paper lays out the engineering trade-offs in more detail.
Across those documents, Google stresses radiation testing, close formation flight control and a steep launch cost learning curve that would have to improve significantly before orbital compute could seriously compete with terrestrial data centers.
SpaceX’s Investor Filing Tempers The Hype
SpaceX’s own pre IPO filing, as reported by Reuters and republished on MarketScreener, throws a little cold water on the loftier talk. The company cautioned investors that initiatives to develop orbital AI compute "are in early stages" and "may not achieve commercial viability."
The filing also underscores how much the idea depends on Starship reaching high launch cadence and robust reusability, both of which remain unproven at scale. For any future launch partnership, that translates into a double bet on engineering reliability and sharp declines in launch costs.
Technical And Orbital Risks
Independent experts are not exactly lining up to call this easy. Beyond launch costs, they highlight orbital congestion, collision risk and the problem of cooling dense compute hardware in a vacuum. Analysts writing for Space.com and in a LiveScience roundup note that tightly packed constellations increase conjunction risk, while both optical links and radiative cooling still need to be proven at hyperscale.
Taken together, those concerns mean regulators, insurers and the orbital operations community will have to be involved if orbital data centers are ever going to move beyond the experiment phase.
What This Could Mean Locally
Closer to home, the research effort and early partnerships could translate into new contracts for machine shops, chip assembly operations and satellite integration firms around the Bay. Google’s prototypes would likely trigger a round of testing, supplier deals and hiring that touches engineering outfits and launch support firms across Northern California.
Hoodline has been tracking orbital compute startups and the supply chain players that spring up around them, a pattern that could repeat if Suncatcher moves forward.
There is still a long road between an academic preprint and a functioning orbital cloud. Formal launch contracts, regulatory approvals and those prototype missions will be the milestones to watch. For now, Google’s research and the Wall Street Journal-led reporting have pulled a fringe engineering concept into the mainstream conversation, and the next real updates are likely to arrive through company statements, regulatory filings, and the prototype launches planned for early 2027.
