🔍 Read the full analysis: Google’s First Data Center Satellite Wrote A Haiku: What AI In Space Can Show on ThorstenMeyerAI.com
Get movie nights 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 small satellite carrying its data-center-class TPU generated and sent down a haiku after launching in October. The test is an early demonstration for Project Suncatcher, not evidence that orbital data centers are commercially practical; power, cost, heat, radiation and satellite networking remain major challenges.
Google says its Project Suncatcher prototype has generated and transmitted an AI-written haiku from orbit, marking an early test of whether the company’s data-center-class TPU chips can run machine-learning workloads in space. The satellite used Google’s Gemma 4 model during a ground-station pass near Svalbard, Norway; the demonstration shows that a small AI task worked, not that orbital data centers are ready for commercial use.
The prototype launched on October 1, 2026, aboard a SpaceX rideshare mission, according to Google’s account and the supplied reporting. Built with satellite company Planet, it carries four TPUs and can generate up to about one kilowatt of power, TechCrunch reported. The available power and thermal limits mean it can operate only one chip at a time, in short bursts.
After roughly three weeks of commissioning, engineers sent Gemma 4 a prompt to write a haiku about space. Google said the satellite returned the response as it passed over Svalbard. The poem read: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” Google called the test its “Hello World” moment and said the satellite was operating as expected.
Google describes the result as the first AI response from a data-center-class TPU in orbit. That claim is narrower than saying AI has never run in space: a satellite carrying an Nvidia H100 launched in 2025, and CNBC reported that its operator later ran and queried Google’s Gemma model in orbit. Google’s distinction is the use of its own TPU and its longer-term proposal for interconnected satellite clusters.
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
What the Haiku Test Establishes
The result is a practical milestone because it moves Project Suncatcher from a research proposal to an operating hardware test. A satellite received a prompt, ran an AI model on a TPU, and sent output back to Earth. That sequence lets engineers begin examining how computing hardware behaves under actual orbital heat and radiation conditions, rather than relying only on ground tests.
The broader idea is to use solar-powered satellites to handle some computing tasks above Earth, potentially drawing on sunlight that is less interrupted than at ground-based facilities. Google’s research argues that solar panels in a suitable orbit could be up to eight times more productive than on Earth. That is a company research estimate, not proof that space-based computing will be cheaper or more efficient overall.
The scale gap is substantial. The prototype’s roughly one-kilowatt maximum is tiny beside the 890 megawatts of nuclear power in a deal Google signed on Earth on October 6, according to the source material. The test matters as evidence of technical possibility, but it does not establish that moving data centers into orbit can meet present-day demand, reduce costs or avoid the impacts of launching and operating large numbers of satellites.
Popular right now
KitchenAid Classic Series 4.5-Quart Tilt-Head Stand Mixer, Onyx Black
Dyson V8 Cyclone Cordless Vacuum
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
As an affiliate, we earn on qualifying purchases.
From Suncatcher Proposal to Orbit
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. Its proposed system uses solar-powered satellites carrying TPUs, flying close together and linked by lasers so they can work as a distributed computing cluster. Google initially described a learning mission with Planet involving two prototype satellites by early 2027. The single-satellite test launched earlier, but it is a much smaller step than demonstrating a connected cluster.
Google’s research sketches a possible arrangement of 81 satellites at about 650 kilometers altitude, grouped within a radius of roughly one kilometer. The paper describes neighboring spacecraft separated by around 100 to 200 meters. Google has reported a ground-based test of an optical transceiver pair reaching 1.6 terabits per second; that bench result is not the same as proving the links will work reliably between satellites in orbit.
Cost is another open question. In Google’s paper, launch costs at current prices work out to about $14,700 per kilowatt per year, compared with about $810 if launch costs fall to $200 per kilogram. The paper says reaching that lower launch-cost assumption would require roughly 1,800 Starship launches. These are modeled estimates tied to stated assumptions, not prices for an operating orbital data center.
““the first-ever AI response from a data center class TPU in orbit.””
Costs and Orbital Risks Remain
The haiku test does not establish how much useful computing the satellite can sustain over time, how reliably its hardware will work through repeated exposure to radiation and thermal stress, or whether the system can be maintained when something fails. The source material says the prototype can run one TPU at a time in bursts of roughly 20 minutes; longer-term performance and operational reliability have not yet been reported.
Economics also remain unsettled. Google’s cost projections depend on future reductions in launch prices, and the paper’s lower-cost scenario would require a very large number of launches. The proposal would also need dependable links to Earth and between spacecraft. The source material cites about 47,000 tracked objects in orbit, but does not specify the measurement date or tracking criteria, so that figure should be treated as the source’s estimate rather than a complete count of orbital objects.
It is also not clear how quickly the prototype will complete the planned tests, what workloads it can handle beyond short demonstrations, or whether Google’s proposed cluster design can operate at its projected scale. The single satellite’s success answers none of those questions by itself.
Thermal Tests and Twin Satellites
Google says the next phase will test how the TPUs respond to thermal stress and radiation in orbit. The company’s next stated milestone is a pair of interconnected satellites in 2027. TechCrunch reported that the twin spacecraft are expected to generate more power and test optical links between satellites.
That mission would be a more direct test of the project’s central networking idea: whether separate spacecraft can exchange data quickly enough to work together on computing tasks. Until those tests are reported, Project Suncatcher remains an early research effort. Google has not established a commercial deployment schedule, and the cost, reliability and environmental implications of a much larger system remain unresolved.
Key Questions
What did Google’s satellite do?
Google says the satellite ran its Gemma 4 model on an orbiting TPU, generated a haiku in response to a prompt and sent the result back to Earth during a pass near Svalbard.
Is this the first AI system to run in orbit?
No. The source material says a satellite carrying an Nvidia H100 launched in 2025, and CNBC later reported that its operator ran and queried Gemma in orbit. Google’s claim is specifically about an AI response from a data-center-class TPU.
Can this satellite power a full data center?
No. The prototype generates up to about one kilowatt and operates one of its four TPUs at a time in short bursts. It is a test satellite, not a full-scale data center.
Why is Google testing AI computing in space?
Google is researching whether solar-powered satellites could support computing with more continuous sunlight and work together through high-speed laser links. Whether that approach can be reliable and cost-effective at scale remains unproven.
What happens after the haiku test?
Google says it will test the TPU’s performance under radiation and thermal stress. Its next stated milestone is a pair of interconnected satellites in 2027 to test optical links and generate more power.
Source: ThorstenMeyerAI.com
