🔍 Read the full analysis: What Google’s Haiku-Writing Satellite Can—and Can’t—Tell Us About AI on ThorstenMeyerAI.com
Get hardware and tech essentials 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
A small Google-Planet satellite running a Google TPU used the Gemma 4 model to answer a prompt with a haiku during a ground-station pass this week. The test shows that an AI workload can run on Google hardware in orbit, but it does not prove that space-based data centers can operate at useful scale or compete on cost.
Google said October 30 that a small satellite running one of its TPU chips in orbit used the Gemma 4 AI model to write a haiku, marking the first reported result from its Project Suncatcher prototype. The demonstration shows that the system can run a basic AI task in space; it does not establish that orbital computing can yet match the capacity, reliability or cost of data centers on Earth.
The satellite, built with spacecraft company Planet, launched on October 1 aboard a SpaceX rideshare mission. Google said it spent about three weeks commissioning the spacecraft and running initial machine-learning tests before engineers sent a prompt during a pass over a ground station near Svalbard, Norway. The model returned: “Blue marble below / Data streams in silent grace / Starlight feeds my core.”
According to hardware details reported by TechCrunch, the satellite carries four TPUs but can power only one at a time. It produces up to about 1 kilowatt, and the chip runs in short bursts; TechCrunch reported bursts of roughly 20 minutes during this week’s test. Google described the haiku as its “Hello World” moment and said the satellite was operating as expected. The test is a demonstration, not evidence of sustained, large-scale computing in orbit.
Google calls the result “the first-ever AI response from a data center class TPU in orbit.” That wording is specific to Google’s chip: a satellite carrying an AI accelerator had already flown. CNBC reported that startup Starcloud launched a satellite with an Nvidia H100 GPU in November 2025 and later ran and queried Google’s Gemma model in orbit. Google’s stated distinction is that its test used a data-center-class TPU, as part of a broader plan to link satellites into a computing cluster.
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
A Prototype, Not an Orbital Data Center
The result matters because it moves Project Suncatcher from research plans to a working, if limited, hardware test. Google is exploring whether solar-powered satellites could eventually provide computing capacity while drawing on sunlight that is more consistently available in certain orbits. The company’s research says solar panels in a suitable orbit could be up to eight times more productive than on Earth. That is a potential advantage, not a measure of the prototype’s actual performance or economics.
Demand for electricity is one reason technology companies are considering new power sources and infrastructure. The source material cites U.S. data centers as using about 4.4% of U.S. electricity in 2023, with the Department of Energy projecting a range of 6.7% to 12% by 2028. Those figures describe electricity use on Earth; they do not show that satellites could replace terrestrial data centers. The test’s roughly one-kilowatt power limit is small beside the 890-megawatt nuclear power agreement Google signed for Earth-based supply on October 6.
For businesses and the public, the near-term takeaway is a technical milestone rather than a new computing service. The demonstration does not show that customers can send workloads to orbit, or that doing so would reduce costs or emissions. Any future effect would depend on whether Google and other companies can solve engineering and launch-cost problems at much larger scale.
Products Worth Considering
Trezor Safe 7 Crypto Hardware Wallet
Ledger Flex + Folio Crypto Hardware Wallet
LG UltraGear 32GX870A-B 32-inch 4K OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Research Paper to First Test
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. Its proposed system uses solar-powered satellites equipped with TPUs, flying close together and communicating through optical, or laser, links. Google’s paper describes clusters of 81 satellites within a radius of about one kilometer at an altitude of roughly 650 kilometers. The company has also reported a bench test reaching 1.6 terabits per second with a pair of optical transceivers; that result was a ground test, not an orbital link demonstrated by this prototype.
Google had previously described a “learning mission” with Planet, with two prototype satellites expected by early 2027. The first prototype launched earlier, on October 1, 2026, and the company reported the haiku test on October 30. Google says the next checks will examine how TPUs handle thermal stress and radiation in orbit. A two-satellite mission planned for 2027 would test interconnected spacecraft and optical links, according to Google and TechCrunch.
The economics remain uncertain. Google’s research estimates that, at current launch prices, putting a kilowatt of computing capacity in orbit would cost about $14,700 per year in launch costs. Its estimate falls to about $810 per kilowatt-year if launch costs reach $200 per kilogram. The research says that lower launch-cost scenario would require roughly 1,800 Starship launches. These are modeled estimates, not costs observed from operating a commercial orbital data center.
““The first-ever AI response from a data center class TPU in orbit.””
The Technical and Cost Questions
The haiku test leaves open whether the hardware can run reliably for extended periods, handle demanding workloads or scale beyond a single satellite. Google has not yet shown that the prototype can sustain the performance required for a large computing cluster. The planned tests of heat and radiation exposure are still ahead, as are the orbital laser-link trials.
Space also brings constraints that the demonstration does not resolve: managing heat in a vacuum, repairing equipment that fails, maintaining dependable connections to Earth and operating amid orbital debris. The source material cites about 47,000 tracked objects already in orbit, but does not give a date or define which kinds of objects are included. The financial estimates depend heavily on future launch prices, and the supplied information does not establish when those prices might be reached. No commercial operating date, customer access plan or full lifecycle cost has been announced.
Thermal Tests and Two-Satellite Links
Google says engineers will continue testing the prototype, including how its TPU hardware responds to thermal stress and radiation. The next stated milestone is a pair of interconnected satellites in 2027. Google’s research plan and reporting by TechCrunch indicate that the mission is intended to test optical links between spacecraft; the pair is also expected to generate more power than the current prototype.
Those tests should provide a clearer answer on whether satellites can exchange data fast enough to work together, and whether the hardware can function in orbit beyond a brief demonstration. They will not, by themselves, settle the questions of launch economics, maintenance or commercial scale. For now, Google has shown that a TPU-equipped satellite can return an AI-generated response—not that orbital data centers are ready to compete with facilities on the ground.
Key Questions
What did Google’s satellite do?
It ran the Gemma 4 model on a Google TPU and returned a haiku after engineers sent a prompt during a ground-station pass near Svalbard, Norway.
Was this the first AI system in orbit?
No. The source material says Starcloud launched a satellite carrying an Nvidia H100 GPU in November 2025 and later ran and queried Gemma in orbit. Google’s claim concerns an AI response from a data-center-class TPU.
Can the satellite run a full data center workload?
The demonstration does not establish that. TechCrunch reported that the satellite carries four TPUs but can power one at a time, with operation in short bursts. Google’s larger cluster concept remains under development.
Why put AI computing in space?
Google is investigating whether solar power and linked satellites could support computing in orbit. Its research estimates that solar panels in a suitable orbit could be up to eight times more productive than on Earth, but that is a potential advantage, not proof that orbital computing is cheaper or practical today.
What is the next milestone?
Google says it will test the current satellite’s response to thermal stress and radiation, then aims to test two interconnected satellites in 2027. The later mission is expected to evaluate optical links between spacecraft.
Source: ThorstenMeyerAI.com
