🔍 Read the full analysis: Google’s Space Satellite Haiku: Evidence, Limits, And AI’s Next Questions on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A Google-Planet prototype satellite sent a haiku generated by Google’s Gemma 4 model running on a TPU, marking an early test of Project Suncatcher. The demonstration confirms a small AI workload ran in orbit; it does not show that orbital data centers are practical or economical. Power, heat, radiation, communications, repairs and launch costs remain major hurdles.
Google says a satellite carrying its TPU AI chip has generated and sent a short poem from orbit using the company’s Gemma 4 model, marking the first reported operating result from Project Suncatcher, its research effort into space-based AI computing. The test shows a small AI workload can run on the prototype; it does not establish that putting large-scale computing infrastructure in orbit is technically or economically viable.
The prototype was built with satellite company Planet and launched on October 1 aboard a SpaceX rideshare mission. After several weeks of commissioning, Google said engineers sent a prompt during a ground-station pass as the satellite went over Svalbard, Norway. The satellite returned a haiku generated by Gemma 4: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” Google described the test as its “Hello World” moment.
According to Google, the spacecraft’s early checks included powering up its host computer and TPU trays and running initial machine-learning tests. A Google spokesperson told TechCrunch the satellite was “operating nominally.” TechCrunch reported that the satellite carries four TPUs but has a power supply of up to about 1 kilowatt, limiting it to one chip at a time and short operating bursts. The company says engineers also tried space-themed limericks and asked the system to report its technical status.
Google calls this the “first-ever AI response from a data center class TPU in orbit.” That wording is specific to its TPU: a separate company, Starcloud, launched an Nvidia H100 GPU in 2025 and later ran Google’s Gemma model in orbit, according to CNBC. Google’s stated distinction is both the chip used and its longer-term plan to connect multiple satellites so they can work together.
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 proof of operation at small scale: a satellite computer received a prompt, ran an AI model and returned its output. That is a meaningful early test for engineers studying whether computing hardware can function in orbit, but the poem itself says little about performance on demanding workloads. Google has not presented this demonstration as a working orbital data center.
The proposal matters because AI computing on Earth consumes substantial electricity. The source material cites U.S. data centers as using about 4.4% of national electricity in 2023, with the Department of Energy projecting a range of 6.7% to 12% by 2028. Those estimates describe U.S. data centers, not Google’s own use. Space could offer strong solar exposure in selected orbits, but moving computation there would introduce costs and engineering problems absent from terrestrial facilities.
Google’s research puts the economics in sharp relief. Its paper estimates orbital power costs of about $14,700 per kilowatt per year at current launch prices, compared with about $810 if launch costs fall to $200 per kilogram. The latter is a scenario, not a price Google has demonstrated. The company’s paper estimates that reaching the lower-cost case would require roughly 1,800 Starship launches. Until costs fall and systems can operate reliably, the satellite is evidence of a test—not a near-term alternative to data centers on Earth.
Products Worth Considering
Jackery Explorer 300 Portable Power Station, 292Wh
Jackery Explorer 1000 v2 Portable Power Station
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
As an affiliate, we earn on qualifying purchases.
From Suncatcher Plan to Orbit
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. Its concept involves solar-powered satellites carrying TPUs, flying in close formations and communicating through optical, or laser, links. Google’s research describes possible clusters of 81 satellites at about 650 kilometers altitude, with spacecraft roughly 100 to 200 meters apart. These are proposed system designs, not the configuration demonstrated by the current prototype.
Google’s initial public milestone was a learning mission with Planet involving two prototype satellites by 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 on that date. The company’s current plan is to test the satellite’s hardware under orbital conditions before attempting a pair of interconnected spacecraft in 2027.
The broader field is not starting with Google. Starcloud’s earlier H100 launch and reported in-orbit model tests show that other teams have also put AI computing hardware into space. Google’s research instead emphasizes linking satellites into a coordinated computing cluster. Its paper reports a bench test reaching 1.6 terabits per second with one pair of optical transceivers; that ground result is not evidence of a functioning multi-satellite network in orbit.
““The first-ever AI response from a data center class TPU in orbit.””
The Hard Limits Still Untested
The haiku test does not establish how well the TPU performs over extended periods, how often it can run, or whether it can handle larger AI workloads. Google says thermal stress and radiation exposure tests are next; their results have not yet been reported. The prototype’s limited power and short operating bursts also leave open how much useful computing a satellite of this type can provide.
Other unresolved issues include removing heat in a vacuum, maintaining reliable links between satellites and ground stations, and repairing or replacing hardware that fails in orbit. Orbital debris is another concern: the source material cites about 47,000 tracked objects in orbit, though it does not specify a date or the tracking category behind that count. Google has not shown that a close-flying satellite cluster can sustain the data rates and coordination needed for large-scale AI work in space.
The cost estimates depend on future launch prices and assumptions in Google’s research. The lower figure of $810 per kilowatt per year is conditional on launch costs falling to $200 per kilogram; it is not the prototype’s measured operating cost. It remains unclear whether launch costs will reach that level, what the full system would cost to build and operate, or how its economics would compare with terrestrial facilities once communications, maintenance and replacement are included.
Thermal Tests and Satellite Links
Google says it will continue testing the TPUs under thermal stress and radiation exposure in orbit. Those checks should provide early evidence about whether the chip can withstand the conditions and operate reliably beyond a brief demonstration. The company has not provided a public schedule or detailed performance targets for each test.
If the prototype tests proceed as planned, Google’s next stated milestone is a pair of interconnected satellites in 2027. TechCrunch reported that the two spacecraft are expected to generate more power and test optical links between satellites. That mission would address a central part of Suncatcher’s proposal: whether separate spacecraft can exchange data quickly enough to act as a coordinated computing system. Until those tests are completed and their results published, the project remains an early research effort.
Key Questions
What did Google’s satellite do?
Google says its satellite ran the Gemma 4 model on a Google TPU and returned a haiku after engineers sent it a prompt during a ground-station pass over Svalbard. The company called this its “Hello World” test.
Was this the first AI computer in orbit?
No. The source material says Starcloud launched an Nvidia H100 GPU in 2025 and later ran Google’s Gemma model in orbit, as reported by CNBC. Google’s “first” claim is limited to an AI response from a data center-class TPU in orbit.
Can this satellite power a full AI data center?
No such capability has been demonstrated. TechCrunch reported that the prototype generates up to about 1 kilowatt and can operate one TPU at a time in short bursts. Google’s proposed larger system would require multiple satellites and further testing.
Why is Google testing AI in space?
Google is investigating whether solar-powered satellites could support computing in orbit, partly against the backdrop of rising electricity demand from data centers. The company’s research says solar panels in a suitable orbit could be more productive than on Earth, but the cost and engineering case remains unproven.
What is the next milestone?
Google says it will test the hardware’s response to heat and radiation. If those tests go well, the company’s next stated step is a pair of interconnected satellites in 2027 to test optical links and a higher-power configuration.
Source: ThorstenMeyerAI.com
