🔍 Read the full analysis: A Haiku From Space, And A Reality Check For AI Claims 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 haiku using Gemma 4 during a pass over Svalbard, marking an early test of Project Suncatcher. The prototype shows that AI inference can run on a chip in orbit, but its limited power and unresolved costs and engineering challenges do not establish that space-based data centers are practical.
Google says its Project Suncatcher prototype generated a haiku with its Gemma 4 AI model during a satellite pass over Svalbard, Norway, demonstrating that a Google TPU can run an AI task in orbit. The test is a small milestone in research into space-based computing, not evidence that orbital data centers are ready to compete with facilities on Earth.
The satellite launched on October 1, 2026, on a SpaceX rideshare mission, according to Google’s account. Built with satellite company Planet, it spent about three weeks undergoing commissioning checks before engineers sent the prompt asking Gemma 4 to write a haiku about space. Google described the response as its “Hello World” moment and said engineers also tested space-themed limericks and asked the payload to report its technical status.
TechCrunch reported that the spacecraft carries four TPUs and can generate up to about 1 kilowatt of power. That power and the satellite’s heat limits mean it can run only one TPU at a time, in short bursts. Google’s spokesperson told TechCrunch that the satellite was “operating nominally.” Google said the next tests will examine how its TPUs perform under thermal stress and radiation exposure.
The demonstration supports a narrow claim: an AI model running on a data-center-class TPU produced a response from orbit. It does not show that a large model was trained there, or that a network of satellites can operate as one high-performance computing cluster. Google’s longer-term concept depends on satellites communicating through laser links; the prototype haiku test alone does not establish that system’s performance.
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 Haiku Is Not a Data Center
The test matters because it moves Google’s space-computing proposal from research toward hardware testing. If a network of solar-powered satellites could eventually handle some computing workloads, it might offer another source of power as demand for AI computing grows. But this satellite’s output is small: its reported maximum of about 1 kilowatt is far below the scale of terrestrial data centers.
The economic case remains unsettled. In its research, Google estimated that launch costs at current prices would put a kilowatt of orbital capacity at about $14,700 a year. Its estimate falls to about $810 a year if launch costs reach $200 per kilogram, a condition the paper treats as a future possibility rather than today’s price. Google’s analysis says reaching that lower-cost scenario would require roughly 1,800 Starship launches. These are modelled estimates, not operating costs measured from a commercial orbital data center.
For readers, the immediate takeaway is a technical demonstration rather than a new computing service. Google has not announced that customers can use orbital TPUs or that the approach will reduce electricity costs. The work is relevant as companies face growing power needs, but its practical value depends on whether the engineering and launch economics can improve enough to support much larger systems.
Products Worth Considering
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
Logitech MX Brio Ultra HD 4K Webcam, Graphite
DJI Osmo Mobile 7P Phone Gimbal with Mic Mini Transmitter, Infinity Black
As an affiliate, we earn on qualifying purchases.
From Research Paper to Orbit
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. Its proposal describes solar-powered satellites carrying TPUs and linked in close formation, with optical connections intended to move data between spacecraft. Google’s research suggests clusters of 81 satellites at about 650 kilometers altitude, with neighboring satellites 100 to 200 meters apart. Those are design goals in the paper, not a deployed constellation.
The project’s schedule has shifted. Google initially described a learning mission with Planet involving two prototype satellites by early 2027. Instead, a first prototype launched in October 2026. Google also published a peer-reviewed version of its research in the journal Joule on the launch date, according to the supplied source material. The next stated milestone is a pair of interconnected satellites in 2027, which would test optical links and generate more power.
Google’s demonstration is not the first reported use of an AI chip or model in orbit. 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 there. Google’s distinction is its claim about a data-center-class TPU response and its proposed linked-satellite architecture—not the first AI activity in space.
““As the satellite passed by Svalbard, Norway, the downlinked poem confirmed success.””
The Costs and Limits Remain
The demonstration does not show how reliably the satellite can run AI workloads over extended periods, or how its hardware will perform after sustained exposure to radiation and temperature swings. Google says those conditions are the subject of upcoming tests. The source material does not provide results from those tests or independently verified measurements of the satellite’s computing performance.
Other obstacles include removing heat in a vacuum, repairing equipment that cannot readily be serviced, maintaining links to ground stations, and managing orbital traffic. The supplied source cites about 47,000 tracked objects in orbit, but does not specify the date or category behind that count. It should not be read as a count of active satellites alone. Google’s proposed tight satellite formations and laser links also remain to be tested in orbit.
It is also unclear when launch costs could approach the assumptions behind Google’s lower-cost estimate, whether a larger constellation could meet the power and cooling needs of AI workloads, or what services such a system might support. The haiku confirms one successful prompt response, not the commercial feasibility of orbital computing.
Testing Before a Twin-Satellite Trial
Google says it will test the prototype’s TPU performance under thermal stress and radiation exposure. The results should help establish whether the chip can continue operating under orbital conditions beyond an initial prompt demonstration. No results or dates for completing those tests were specified in the supplied material.
If the tests support further development, Google’s next stated milestone is a pair of interconnected satellites in 2027. TechCrunch reported that the follow-up spacecraft are expected to generate more power and test optical links between satellites. That mission would address a central question in Suncatcher: whether separate spacecraft can exchange data fast enough to work together on computing tasks. Until that is tested, the project remains an early experiment rather than an orbital data-center service.
Key Questions
What did Google’s satellite do?
Google says a satellite carrying a TPU ran Gemma 4 to answer a prompt with a haiku during a pass over Svalbard, Norway. The company called it an early “Hello World” test.
Is this the first AI system in orbit?
No. The supplied reporting says Starcloud launched a satellite with an Nvidia H100 in 2025 and later ran Google’s Gemma model there. Google’s narrower claim concerns a response from a data-center-class TPU in orbit.
Can the satellite run a full data center?
No such capability has been demonstrated. TechCrunch reported that the prototype can generate up to about 1 kilowatt and can run only one of its four TPUs at a time in short bursts.
What is the next step for Project Suncatcher?
Google says it will test the chip’s response to radiation and thermal stress. If development proceeds, the company’s next stated milestone is a pair of interconnected satellites in 2027 to test links between spacecraft.
Source: ThorstenMeyerAI.com
