🔍 Read the full analysis: How To Read The Claims Behind Google’s AI Satellite Haiku 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 small satellite built with Planet has run a Google TPU in orbit and returned a haiku generated by Gemma 4, Google said on October 30, 2026. The test shows that a data-center-class TPU can operate in space, but it does not establish that orbital AI computing is practical or economical at scale.
Google says a small satellite carrying its Tensor Processing Units (TPUs) generated and returned a haiku while passing over Svalbard, Norway, in what the company describes as the first response from a data-center-class TPU in orbit. The October 30 test is an early milestone for Project Suncatcher, Google’s plan to study whether solar-powered satellites could eventually support AI computing in space; it does not show that a space-based data center is ready for commercial use.
The satellite, built with Earth-imaging company Planet, reached orbit on October 1 aboard a SpaceX rideshare mission. Google said it confirmed contact that day and that the spacecraft was operating as expected. After about three weeks of commissioning, engineers sent a prompt asking the onboard system to write a haiku about space. Google reported that the response was downlinked during a ground-station pass over Svalbard: “Blue marble below / Data streams in silent grace / Starlight feeds my core.”
The model used was Gemma 4, Google’s open model, running on a Google TPU. Google said engineers also tried space-themed limericks and asked the payload to report its technical status. A Google spokesperson told TechCrunch the satellite was “operating nominally.” These are demonstrations reported by Google and the company’s spokesperson; the available account does not provide independent test results or a detailed performance assessment.
TechCrunch reported that the satellite carries four TPUs but can generate up to about one kilowatt of power. Power and heat limits mean it can operate only one chip at a time, in bursts. TechCrunch described bursts of roughly 15 minutes before launch and about 20 minutes during the reported test. That is a small-scale experiment, not evidence of sustained computing capacity comparable with a terrestrial data center.
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 Orbital Test Can Show
The result matters because it demonstrates a basic operation required by Google’s wider proposal: a TPU can be powered in orbit, run a model, receive a prompt and return an output. Google’s longer-term concept is not simply to place a chip on one satellite. Its research describes clusters of solar-powered satellites, linked by lasers, that could coordinate work more like a computing system on the ground.
Google’s argument is that solar power may be more productive in some orbits than on Earth. The company’s research says a solar panel in a suitable orbit could produce up to eight times as much energy as on Earth. That is a company research estimate, not proof that orbital computing is cheaper overall. Launch, spacecraft construction, thermal control, communications and replacement costs all affect the economics.
The potential relevance is tied to rising electricity demand from computing. The source material cites U.S. data centers using about 4.4% of national electricity in 2023 and a Department of Energy estimate of 6.7% to 12% by 2028. Moving some computing off Earth could, in principle, change where energy demand falls. But the satellite’s roughly one-kilowatt limit shows how far this prototype is from the capacity needed to make a material difference to data-center electricity use.
Google’s own cost estimates underline the gap between a successful test and a viable service. Its research puts the annual launch cost of one kilowatt of orbital capacity at about $14,700 at current launch prices, compared with about $810 if launch costs fall to $200 per kilogram. The lower figure depends on a major decline in launch costs; it is not the price of the current system or a confirmed future operating cost.
Products Worth Considering
Anker SOLIX C300 Portable Power Station, 288Wh
Jackery Explorer 500 v2 Portable Power Station, 512Wh, 500W
Apple MacBook Air 13-inch with M5 chip, 16GB memory, 1TB SSD, and 12MP Center Stage camera (Midnight)
As an affiliate, we earn on qualifying purchases.
From Suncatcher Plan to Prototype
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. The proposal centered on satellites carrying TPUs, flying close together and communicating through optical links. Google initially described a learning mission with Planet involving two prototype satellites, with a target of early 2027. The first prototype launched earlier than that target, but the account provided describes a limited, single-satellite test rather than the planned interconnected system.
Google published a peer-reviewed version of its research in the journal Joule on October 1, 2026, the day the satellite launched, according to the source material. The research sketches clusters of 81 satellites at an altitude of about 650 kilometers, within a radius of roughly one kilometer and separated by about 100 to 200 meters. It also reports a bench test reaching 1.6 terabits per second with a pair of optical transceivers. A bench test is not the same as demonstrating that a full satellite cluster can maintain those links in orbit.
The phrase “first AI in orbit” would be inaccurate based on the supplied reporting. 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 narrower claim concerns a response from a data-center-class TPU. The claimed distinction is the chip and the proposed linked-cluster architecture, not the first use of AI hardware or a model in space.
““the first-ever AI response from a data center class TPU in orbit””
Costs and Space Risks Remain
The haiku test does not establish how reliably the TPU can perform longer or more demanding workloads. Google’s next stated tests concern thermal stress and radiation exposure, two conditions that can affect hardware in orbit. The supplied material does not give results from those tests, nor does it say how often the satellite can run workloads over longer periods.
Several practical questions remain open: how equipment would be repaired or replaced, how heat would be managed in a vacuum, and how much capacity could be delivered over links between satellites and ground stations. The source also cites about 47,000 tracked objects in orbit, but gives no measurement of the collision risk to this particular satellite or a planned cluster. That count should be treated as context, not as a prediction of an incident.
The economics are also unsettled. Google’s lower cost estimate depends on launch prices falling to $200 per kilogram; the source says reaching that scenario would require roughly 1,800 Starship launches. The figure is part of the research’s projected scale, not a schedule or confirmed launch plan. It remains unclear when, or whether, orbital computing could compete with terrestrial data centers after including spacecraft, networking and operating costs.
Thermal Tests and Twin Satellites
Google says the prototype will undergo tests of TPU performance under thermal stress and radiation exposure. If those tests go well, 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 address a different question from the haiku demonstration: whether two spacecraft can exchange data reliably enough to coordinate computing, rather than whether one satellite can run a short prompt. The outcome, including test results, actual power available and link performance in orbit, will help clarify how closely the system can match Google’s proposed clusters. No commercial deployment date or confirmed large-scale operating plan is provided in the source material.
Key Questions
What did Google’s satellite do?
Google said a satellite carrying a TPU ran Gemma 4 and returned a haiku after engineers sent it a prompt. The company described the event as its first AI response from a data-center-class TPU in orbit.
Was this the first AI system in space?
No. The supplied reporting says Starcloud launched a satellite with an Nvidia H100 GPU in 2025 and later ran and queried Google’s Gemma model in orbit. Google’s claim is specifically about a response from its TPU hardware.
Can the satellite run a full data center?
No such capability has been shown. TechCrunch reported that the prototype produces up to about one kilowatt and can operate only one of its four TPUs at a time, in bursts.
Why is Google testing AI computing in orbit?
Google’s research argues that solar panels in a suitable orbit could produce more energy than panels on Earth. Whether that advantage offsets launch, hardware, cooling and communications costs remains unproven.
What is the next milestone?
Google says it will test the satellite under thermal stress and radiation exposure. If those tests go well, it has named two interconnected satellites in 2027 as the next milestone.
Source: ThorstenMeyerAI.com
