Google Project Suncatcher: TPUs Head to Orbit Oct 1

§ 01What Google confirmed#

Google’s post, dated September 24, 2026 and written by Travis Beals, Senior Director, Paradigms of Intelligence, opens with the news: “After years of research, Project Suncatcher is scheduled to embark on its first test in orbit, launching a prototype satellite to evaluate how Google Tensor Processing Units (TPUs) perform in space.” The ride: “This initial mission onboard the upcoming Transporter-18 rideshare mission with SpaceX was developed in partnership with Planet.” The timing: “Putting our first TPUs in orbit next week will help us get data and learnings to inform future launches.”

The New York Times reported it first, as The Verge credits in its 14:15 UTC story. Ars Technica followed at 16:16 UTC with the date, October 1, and the spacecraft’s details.

01Three things space does to a chip, and what Google measured on the ground

02From a 2025 paper to a 2026 orbit, and the two-satellite test that comes next

1 / 2

←→

§ 02What the satellite tests#

Google frames the whole mission as one question: “Can our AI hardware operate in space?” Its stated goal: “It’s designed to gather in-orbit data on how our TPUs handle the physical stress of spaceflight and the radiation and thermal extremes of space.”

Launch. Google says a trip to low Earth orbit brings “sustained acceleration loads up to 10 times the force of gravity”, and that chips can see “even greater forces up to 50 to 100 g”. The team shook the satellite on all three axes and, in Google’s words, “we were pleasantly surprised that the hardware held up to the force.”

Radiation. Google tested TPUs in a proton beam at UC Davis’s Crocker Nuclear Laboratory while they ran AI workloads, and reports: “Initial results have shown that our Trillium TPUs hold up remarkably well, and can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission.” The numbers behind that line are in Google Research’s 2025 paper summary.

Cooling. “But in space, there’s no airflow,” Google writes; heat can only leave through radiators, so the team uses “a combination of heat pipes and radiators to cool the chips.” Ars reports the design as a thermal interface material feeding aluminum and copper heat pipes, and that the system can only keep up for about 15 minutes before the TPUs shut down to let the radiators catch up.

§ 03What it does not test yet#

The Suncatcher idea is not one satellite. Google’s end state is clusters of satellites, each with dozens of TPUs, talking to each other over lasers at very high bandwidth across very short distances. Google compares the pointing problem to “hitting a coin-size target from miles away while both points are in motion”, and says: “We’ll test our work on this in 2027 when we put two satellites in orbit.”

That 2027 flight was the original first step. When Google announced Suncatcher on November 4, 2025, the plan was “to launch two prototype satellites by early 2027”, and Planet described the same pair flying in tandem with cross-links. Per Ars, Google wanted to move faster, so it fitted its chips into a satellite Planet had already built. The 2027 launches are still planned.

§ 04Why Google wants chips in orbit#

Google’s answer is power: in low Earth orbit, satellites “can access near-constant sunlight, generating up to eight times more solar power than on Earth.” The 2025 paper models an illustrative 81-satellite cluster at a mean altitude of 650 km, and argues that if launch prices keep falling, “prices may fall to less than $200/kg by the mid-2030s”, the point at which launching and operating a space data center could roughly compare with the energy costs of one on Earth.

That is a long bet, and Google writes it as one: “This first launch is about seeing what works, identifying points of failure, and applying those findings to future missions.” Ars adds that the team expects years before Suncatcher moves from project to product.

§ 05What it means if you run AI on real work#

Nothing about your workload moves to orbit. The Gemini models behind Gemini 3.8 Flash, which runs CellCog’s Flash-tier agents, are served from data centers on the ground, and so is everything else a business uses today. Suncatcher is Google testing whether the power that runs AI a decade from now could come from sunlight above the atmosphere rather than from a grid that is already fighting over data centers.

What a business can use now is the thing those chips run: models good enough to do real work, and a harness that points them at your tasks, checks the results and asks before anything leaves your walls. The Gemini 4 tracker follows the next model Google says is on its way.

§ 06What we are watching for#

• The launch itself, October 1, and whether Transporter-18 holds its date.
• Google’s first in-orbit results: whether the TPUs, the cooling bursts and the radiation numbers hold up outside the lab.
• The two-satellite laser-link test in 2027, the flight that tests the actual Suncatcher architecture.
• Planet’s side of the mission, and whether its Owl satellite bus carries the 2027 pair.

§ 07The record#

As of September 24, 2026, 18:30 UTC: page opened. Google’s post is dated September 24 with no time shown; The Verge published at 14:15 UTC crediting The New York Times; an X post from @ns123abc spread the news at 15:22 UTC; Ars Technica published at 16:16 UTC with the October 1 date. The New York Times article itself was not read for this page; its figures appear here only as Ars Technica and The Verge report them.