Google just put its AI chips in orbit aboard a SpaceX rocket, kicking off a project that demands 1,800 Starship launches over the next decade — a pace only Elon Musk's outfit has any shot at hitting, and one that proves American builders move faster when bureaucrats can't touch them.

The test, called Project Suncatcher, is Google's opening bid to move data centers off-world — where land-use fights, grid constraints, and regulatory friction don't exist. If it works, orbital compute clusters powered by uninterrupted solar could outcompete anything strangled by Earth-bound permitting agencies. The question is whether the launch math pencils out.

On October 1, a SpaceX rocket carried Google's experimental satellite — built by Planet Labs and loaded with Tensor Processing Units — into low-Earth orbit. The chips, which compete with Nvidia's GPUs for AI workloads, will fire in 15-minute bursts to test whether they can survive the hostile space environment without cooking themselves.

"We've done testing on the ground, but you know, there's no test that's completely as good as the real thing," said Travis Beals, the Google executive managing Project Suncatcher.

Ground tests at a UC Davis proton beam facility simulated solar radiation and cosmic rays. Google says the chips held up "remarkably well" and could handle more ionizing radiation than they'd be expected to encounter over five years in orbit. A new cooling system had to be built from scratch — there's no airflow in space to disperse the intense heat TPUs generate.

Google also released a peer-reviewed white paper in the journal Joule laying out the economics. The upshot: getting launch costs down to roughly $200 per kilogram by 2035 would require Starship to haul 370,000 tons of payload into orbit. That translates to roughly 1,800 launches over ten years — 180 per year, each carrying 200 metric tons.

Starship has never flown more than five times in a year. SpaceX says it will fly far more than that. Nobody else is even close.

Google is also a major investor in SpaceX, which raises the obvious question: is this rigorous analysis or a friendly plug for a portfolio company? The researchers stress it's not an economic feasibility study — though they project a 20 percent annual price-reduction learning curve based on SpaceX's track record since Falcon 1.

Scientific American framed the story around engineering challenges and astronomer complaints about light pollution from orbital infrastructure. TechCrunch focused on the launch math and Google's long-term bet. What neither outlet dwells on is the real reason Big Tech is looking up: terrestrial data centers are "increasingly politically unpopular," as Scientific American put it — a polite way of saying communities and regulators are choking expansion on the ground.

OpenAI and xAI are pursuing orbital data centers too. The race is on to escape the permitting purgatory that makes building anything major in America a decade-long ordeal.

Google plans a second satellite in 2027 to test laser communications between orbital nodes. The full vision: 81 satellites flying in close formation, processing in parallel — no grid hearings, no environmental impact statements, no zoning boards.

The math is brutal — 1,800 Starship launches in a decade is a staggering ask for a vehicle still proving its cadence. But if Musk hits the number, the payoff isn't just cheaper AI compute. It's proof that the future belongs to whoever builds fast enough to leave the gatekeepers behind.