Google's Suncatcher prototype is in orbit: what it is testing

Google said on 1 October 2026 that the prototype satellite for Project Suncatcher, its research effort on running AI hardware in orbit, has launched aboard SpaceX's Transporter-18 rideshare mission. In the post, Google says its team "has confirmed contact with the satellite and it is operating as expected".

What Project Suncatcher is

Google announced Suncatcher on 4 November 2025 as "our new research moonshot to one day scale machine learning in space". The idea, in Google's words, is to explore "how an interconnected network of solar-powered satellites, equipped with our Tensor Processing Unit (TPU) AI chips, could harness the full power of the Sun".

On our reading, power is the business angle. In a 24 September 2026 post, Google said that in low Earth orbit "satellites can access near-constant sunlight, generating up to eight times more solar power than on Earth". The launch-day post calls the whole programme "a long-term research moonshot exploring whether space could one day host scalable machine learning infrastructure".

What the prototype is testing

The satellite was built with the satellite company Planet. Google says that "over the coming weeks" it will "gather in-orbit data on how our TPUs handle the physical stress of spaceflight and the radiation and thermal extremes of space".

Google's 24 September post covers these engineering problems:

• Launch stress. Google says it shook the satellite on all three axes to mimic a rocket launch, and "we were pleasantly surprised that the hardware held up to the force".
• Radiation. Google says it tested TPUs in a proton beam at UC Davis's Crocker Nuclear Laboratory while running AI workloads, and that "initial results have shown that our Trillium TPUs hold up remarkably well".
• Cooling. "In a vacuum, you can only diffuse heat via radiators," the post says. Google says it has tested heat pipes and radiators in a thermal vacuum chamber: "We'll see how our new TPU cooling system works in space."

The same post says future satellites would fly in clusters and talk to each other by laser, and that "we'll test our work on this in 2027 when we put two satellites in orbit".

What the paper says

Google says a peer-reviewed paper is "now available in Joule". We could not open the publisher's page, so we have gone by the paper's published abstract, which we read through a research index, and by the earlier version of the paper that Google posted in 2025. According to the abstract, the paper:

• describes fleets of satellites with solar arrays, laser links between satellites (free-space optics) and Google TPU chips, flown close together;
• illustrates formation flying with an 81-satellite cluster of 1 km radius;
• reports that Trillium TPUs survived a radiation dose equivalent to a five-year mission without permanent failures;
• treats launch as critical to overall system cost, and says a learning-curve analysis suggests launch to low Earth orbit may reach $200/kg or less by the mid-2030s.

That launch figure is a projection in the paper, not a current price.

What is not shown yet

• Results from orbit. Google's post reports contact and that the satellite is "operating as expected". It does not report TPU performance in space; it says it will share more "as the mission unfolds".
• Links between satellites. Google says these will be tested in 2027 with two satellites. When it announced Suncatcher in 2025, Google said its next step was to "launch two prototype satellites by early 2027". Google has not said how that two-satellite plan maps onto the current one.
• Cost. None of the three Google posts we read gives a cost for computing in orbit.

Why it matters

On our reading, the pressure behind projects like this is power. As we covered in our 25 September Daily Update on the UK's 'largest AI supercomputer', which could wait until the 2030s for power, the Guardian reported that UK Power Networks has reportedly told Nscale the grid will not be able to supply enough power to its Loughton datacentre until the early to mid-2030s. The Guardian also cited Ofgem figures from June showing 315 datacentres queueing to connect to the National Grid.

Google calls this launch the first step in a long-term research moonshot. On our reading, it does nothing for grid queues today.