🔍 Read the full analysis: What Google’s Satellite Poetry Experiment Means For Artificial Intelligence on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Google says a satellite carrying its TPU chip generated a short poem using Gemma 4 during a pass over Svalbard. The test is an early milestone for Project Suncatcher, but the prototype is small, and the cost, cooling, radiation and networking challenges of orbital AI remain unresolved.
Google says its Gemma 4 AI model generated a poem aboard a satellite carrying a Google TPU, marking the first reported prompt response from the chip in orbit. The test, conducted during a pass over Svalbard, Norway, is an early milestone for Project Suncatcher, the company’s effort to study whether groups of solar-powered satellites could one day perform AI computing in space.
Google said engineers sent a prompt to the satellite and received a three-line response: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” The company called the exchange its “Hello World” moment. The satellite had been in orbit for several weeks while engineers completed initial checks and machine-learning tests.
The prototype launched on October 1 aboard a SpaceX rideshare mission, alongside more than 100 other spacecraft, according to the source material. Built with satellite company Planet, it uses a standard Planet platform and carries four TPUs. TechCrunch reported it can generate up to about one kilowatt of power and can operate only one chip at a time. Google said the spacecraft was operating as expected; a spokesperson told TechCrunch it was “operating nominally.”
Google’s next tests are intended to assess how the chips perform under thermal stress and radiation. The company’s proposed longer-term system would use close-flying satellites linked by lasers. Google’s research describes clusters of up to 81 satellites, but that network has not been demonstrated in orbit. The successful poem is evidence of a limited computing task working on one prototype, not proof that an orbital data center is practical at commercial scale.
Infographic 01 · What flew and what it did
One satellite, four chips, one haiku
Google’s Project Suncatcher prototype is the first test of whether its AI data center chips can work in orbit. Here is what is on board and how it got there.
- 4 Nov 2025Project Suncatcher announcedResearch post and preprint; plan: two prototypes with Planet by early 2027
- 1 Oct 2026Prototype launchedSpaceX Transporter-18 rideshare, built with Planet; peer-reviewed paper in Joule
- Oct 2026About three weeks of commissioningHost and TPU trays powered up, early machine learning tests
- Week of 26 Oct“Hello World”First prompt to Gemma 4; the haiku came down during a pass over Svalbard, Norway
- NextHeat and radiation tests in orbitHow the TPUs behave under thermal stress and real space radiation
- 2027Twin interconnected satellitesMore power and laser links between satellites (Google, TechCrunch)
Scale check
The satellite vs. one power deal on Earth
890 MW = 890,000 kW, about 890,000 times the prototype (my arithmetic). The prototype bar is drawn at a visible minimum; to scale it would be invisible.
Not the first AI chip in orbit
Starcloud-1, launched November 2025 (CNBC)
- Carried one Nvidia H100, a data center GPU
- Trained a small language model in orbit
- Ran Google’s open model Gemma up there in December 2025
- Google’s claim is narrower: the first response from a data center class TPU in orbit
Sources: Google, 30 Oct 2026 · Google, 1 Oct 2026 · Google Research, 4 Nov 2025 · TechCrunch, 30 Oct 2026 · TechCrunch, 1 Oct 2026 · Google and Constellation, 6 Oct 2026 · CNBC, 10 Dec 2025.
Powered by Thorsten Meyer AI
Why Orbital AI Faces a Cost Test
The experiment addresses a real pressure point for AI companies: data centers need large amounts of electricity. The source material cites U.S. Department of Energy estimates that data centers used about 4.4% of U.S. electricity in 2023, with the share projected to reach between 6.7% and 12% by 2028. Moving some computing off Earth could, in principle, tap solar power more directly. Google’s research says panels in a suitable orbit could be up to eight times more productive than on Earth.
But access to sunlight is only one part of the equation. In its research, Google estimated that launch costs at today’s prices would make a kilowatt of orbital computing capacity cost about $14,700 per year. The paper’s estimate falls to about $810 per kilowatt-year if launch prices reach $200 per kilogram, a level Google compares with the cost of power for U.S. data centers. The lower figure depends on future launch prices and is not the operating cost established by this prototype.
Google’s paper also sets out the scale needed to make its proposed approach work economically: roughly 1,800 Starship launches, according to the source material. That makes the project relevant to the future of AI infrastructure, but also underlines the gap between a successful demonstration and a viable business. For now, the test shows that a small satellite can run a model and return an answer; it does not show that doing so is cheaper, more reliable or easier than operating computing equipment on Earth.
Products Worth Considering
Bose QuietComfort Ultra Wireless Noise-Cancelling Headphones with Spatial Audio, Black
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
Logitech MX Brio Ultra HD 4K Webcam, Graphite
As an affiliate, we earn on qualifying purchases.
From Suncatcher Proposal to Orbit
Google introduced Project Suncatcher on November 4, 2025, in a research post and a preprint paper. It outlined an idea for solar-powered satellites carrying TPUs, flying close together and exchanging data over optical links. Google initially described a learning mission with Planet and two prototype satellites as a milestone for early 2027.
The first prototype reached orbit earlier, on October 1, 2026. Google also published a peer-reviewed version of its research in the journal Joule that day, according to the supplied material. After commissioning, engineers tested the satellite’s computer and TPU hardware before sending the prompt that produced the poem on October 30.
The test is not the first time an AI chip or model has operated in space. CNBC reported that startup Starcloud launched a satellite carrying an Nvidia H100 GPU in November 2025 and later ran and queried Google’s Gemma model in orbit. Google’s stated distinction is narrower: it calls this the first AI response from a data-center-class TPU in orbit. Its proposed satellite cluster, connected to run computing workloads together, is a separate part of the Suncatcher plan that remains to be tested.
““the first-ever AI response from a data center class TPU in orbit””
The Unsolved Limits of Orbital Computing
The demonstration does not establish how the prototype performs on sustained or demanding workloads. Its reported power supply limits it to one TPU at a time, and the source material describes operation in bursts of about 20 minutes. It is not clear from the available information how much usable computing the satellite can provide over time, or how its performance compares with a terrestrial system running the same workload.
Several engineering and operating questions remain open: how equipment can shed heat in a vacuum, how chips will withstand radiation, how satellites could be repaired, and how reliably they can communicate with ground stations. The proposed laser links between satellites also need an in-orbit demonstration. Google has reported a ground-based bench test reaching 1.6 terabits per second between a pair of optical transceivers, but that is not evidence of the same performance in space.
Cost is another uncertainty. Google’s estimates depend on launch prices falling substantially, while the source material gives no confirmed schedule or cost for a commercial system. It also cites about 47,000 tracked objects already in orbit, a relevant congestion concern, but does not establish how much additional traffic Suncatcher would create. The project remains an experiment rather than a deployed service.
Twin-Satellite Tests in 2027
Google says it will continue testing the current spacecraft, including the TPUs’ response to radiation and thermal conditions. If those tests go well, the company’s next stated milestone is a pair of interconnected satellites in 2027. TechCrunch reported that the pair is expected to generate more power and test optical links between spacecraft.
That mission would address a key question for Suncatcher: whether satellites can exchange data fast enough to work together on computing tasks. Google has not, in the material provided, announced a date for a commercial service or set out a firm deployment schedule beyond the next prototype milestone. Until the linked-satellite test and further cost and reliability evidence arrive, the poem remains a small but concrete demonstration of computing in orbit—not a working space-based data center network.
Key Questions
What did Google’s satellite do?
Google says its satellite ran Gemma 4 on a TPU and returned a poem after engineers sent it a prompt during a pass over Svalbard. The company described the response as its “Hello World” test.
Is this the first AI in space?
No. The supplied material says Starcloud launched a satellite with an Nvidia H100 GPU in 2025 and later ran Google’s Gemma model in orbit. Google’s claim is specifically about a response from a data-center-class TPU in orbit.
Can the satellite run a full data center?
No. The prototype carries four TPUs but reportedly has up to about one kilowatt of power and can run one chip at a time. It is a test platform, not a full-scale data center.
What is Google testing next?
Google says it will test how the TPU hardware handles radiation and thermal stress. The company has also identified two interconnected satellites in 2027 as its next milestone, subject to the current tests.
What is the main barrier to orbital AI?
There is no single settled answer. Google’s research points to launch cost, while engineers must also address heat management, radiation, repairs, communications and the challenge of linking satellites into a functioning computing cluster.
Source: ThorstenMeyerAI.com
