🔍 Read the full analysis: Can AI Create In Orbit? What Google’s Haiku-Writing Satellite Reveals 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- and Planet-built satellite carrying four TPUs received a prompt and returned a Gemma 4 haiku during a pass over Svalbard, Google said on October 30, 2026. The test shows that a data-center-class TPU can run briefly in orbit; it does not establish that space-based AI computing is practical or economical at scale.
Google says a satellite carrying its TPU chip has generated and sent back a haiku using its Gemma 4 AI model, marking the first reported response from the company’s Project Suncatcher prototype in orbit. The test, conducted during a pass over Svalbard and reported October 30, shows that the hardware can run an AI task in space, but it does not demonstrate a commercially viable orbital data center.
The prototype, built with satellite company Planet, launched on October 1 aboard a SpaceX rideshare mission. After several weeks of commissioning, engineers sent a prompt asking Gemma 4 to write a haiku about space. Google said the satellite returned: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” The company described the exchange as its “Hello World” moment.
According to hardware details reported by TechCrunch, the satellite uses a standard Planet platform and carries four Google TPUs. It produces up to about one kilowatt of power, limiting operation to one chip at a time and short bursts. TechCrunch reported that the test lasted about 20 minutes. Google said the spacecraft was “operating nominally” and that early tests included powering up the computer and running machine-learning tasks.
The haiku is a demonstration of a narrow technical capability, not evidence that a full-scale AI workload can be operated in orbit. Google’s next tests are intended to examine how the chips handle radiation and thermal stress. The company has described a later test involving two connected satellites, planned for 2027, to examine links between spacecraft.
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 Small Test of Orbital Computing
The experiment matters because it moves Google’s proposal for solar-powered AI computing in space from research into an on-orbit hardware test. AI data centers on Earth require large amounts of electricity, and Google’s research argues that solar panels in a suitable orbit could produce substantially more energy than comparable panels on the ground. That is a potential benefit, not a demonstrated cost advantage: this satellite’s brief, low-power operation is far from the scale required for today’s large AI systems.
Cost is a central constraint. Google’s research estimates annual launch costs of about $14,700 per kilowatt at current launch prices, compared with roughly $810 per kilowatt if launch costs fall to $200 per kilogram. Those are estimates based on stated launch assumptions, not the operating cost of a proven orbital data center. The research suggests that matching terrestrial power costs would depend on a major reduction in launch prices and a much larger deployment.
The test also offers a practical way to study whether chips can keep working amid radiation, heat-management challenges and limited opportunities for repair. For companies considering space-based computing, the result is a sign that basic operations are possible, while the engineering and economic hurdles remain substantial. For the public, it is an early experiment rather than an imminent shift in where everyday AI services run.
Products Worth Considering
Samsung 27-Inch Odyssey QD-OLED G8 4K 240Hz Gaming Monitor
Anker SOLIX C300 Portable Power Station, 288Wh
Samsung Galaxy Tab S10 FE 128GB Wi-Fi Android Tablet, Blue
As an affiliate, we earn on qualifying purchases.
From Research Plan to Orbit
Google introduced Project Suncatcher in November 2025, describing a longer-term concept: closely spaced satellites carrying TPUs and using optical links to operate as a connected computing cluster. Its initial public plan called for a learning mission with Planet and two prototype satellites by early 2027. The single satellite launched in October 2026 is an earlier, smaller step than that proposed two-spacecraft test.
Google is not the first company to put an AI-capable chip in orbit. CNBC reported that startup Starcloud launched a satellite carrying an Nvidia H100 in November 2025 and later ran and queried Google’s Gemma model there. Google’s distinction is narrower: it says this was the first response from a data-center-class TPU in orbit. Its longer-term proposal also emphasizes linking multiple satellites so they can work together, rather than running tasks on separate spacecraft.
In its research, Google sketches clusters of 81 satellites at about 650 kilometers altitude, flying close enough for high-speed optical links. The paper reports a laboratory test of a link between two transceivers; that is different from demonstrating a working orbital cluster. The prototype’s haiku test did not establish that those links, cluster-scale computing or sustained operation have been achieved.
““Blue marble below / Data streams in silent grace / Starlight feeds my core.””
— Google, describing the returned poem
The Limits Still Untested
The public account does not give detailed results for the satellite’s AI performance, including throughput, reliability over longer periods or comparisons with equivalent ground-based hardware. It also does not establish how much energy the complete system uses across all components, or what the prototype cost to build and operate. The reported 20-minute operating window is a brief demonstration, not evidence of continuous computing capacity.
Key engineering questions remain open: how the system will shed heat in a vacuum, withstand radiation over time, receive repairs or replacement parts, and communicate reliably with the ground. A larger constellation would also have to contend with orbital debris; the source material cites about 47,000 tracked objects, without specifying a measurement date or defining which objects are included. Google has not shown that its proposed multi-satellite links work in orbit or that launch costs will fall to the levels in its economic estimates.
Radiation Tests and Twin Satellites
Google says it will continue testing the current satellite, including its TPU trays under thermal stress and radiation exposure. The company’s next stated milestone is a pair of interconnected satellites in 2027. According to the source material, that mission is expected to produce more power and test optical links between spacecraft.
Those results will help show whether Google can move beyond a short AI response toward sustained workloads and satellite-to-satellite computing. The timing and specifications of the planned mission, the performance targets for the tests and the costs of any larger deployment have not been detailed in the material available. For now, the confirmed development is limited: a TPU in one satellite ran a prompt and returned a poem.
Key Questions
What did Google’s satellite do?
Google said its satellite received a prompt, ran the Gemma 4 model on a Google TPU and sent back a haiku during a pass over Svalbard.
Was this the first AI system in orbit?
No. The source material reports that Starcloud launched a satellite with an Nvidia H100 in 2025 and later ran Google’s Gemma model there. Google’s narrower claim is that its satellite returned the first response from a data-center-class TPU in orbit.
Can the satellite run a full AI data center?
No such capability has been demonstrated. The prototype produces up to about one kilowatt and operates one TPU at a time in short bursts. Google’s proposed connected satellite clusters remain a future development.
Why put AI computing in space?
Google’s research argues that solar panels in some orbits could generate more power than panels on Earth. Whether that benefit can outweigh launch, hardware, communications and maintenance costs at useful scale remains unproven.
What is the next milestone?
Google plans further tests of the current TPU under radiation and thermal stress, followed by a stated goal of testing two interconnected satellites in 2027. The planned mission’s full specifications have not been provided here.
Source: ThorstenMeyerAI.com
