🔍 Read the full analysis: The Meaning Of Google’s Satellite Haiku For AI In Space on ThorstenMeyerAI.com
Get smart everyday buys 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 Planet-built satellite carrying Google TPUs returned a haiku generated by Gemma 4 during a ground-station pass over Svalbard, Google said. The test shows that a data-center-class TPU can run in orbit, but the satellite is a small prototype, not an orbital data center; power, heat, radiation, repair and launch costs remain major barriers.
Google said a satellite carrying its TPU AI chip generated a three-line poem with its Gemma 4 model and sent it to a ground station during a pass over Svalbard, Norway. The test, reported October 30, is an early result for Project Suncatcher, Google’s effort to study whether solar-powered satellites could eventually host AI computing; it does not show that an orbital data center is commercially viable.
The prototype, built with satellite company Planet, launched October 1 on a SpaceX rideshare mission. Google said the spacecraft completed commissioning checks, including powering up its host computer and TPU hardware, before engineers sent prompts and ran initial machine-learning tests. The haiku was a deliberately simple “Hello World” check: Google described the returned text as confirmation that the system could receive a prompt, run an AI model and send a response back to Earth.
According to TechCrunch, the satellite carries four TPUs and can generate up to about one kilowatt of power. That limits it to running one chip at a time and in short bursts; the report described runs of about 20 minutes during this week’s test. Google said the satellite was operating as expected, but did not publish detailed performance results, power use for the poem or a measure of how its in-orbit computing compares with the same hardware on the ground.
The result is narrower than a claim that Google has put a working data center in orbit. A satellite carrying an Nvidia H100 GPU launched in November 2025, and CNBC reported that its operator, Starcloud, later ran a small language model and Google’s Gemma model in space. Google’s specific claim is that its test produced the first AI response from a data-center-class TPU in orbit. The larger Suncatcher concept would link satellites into computing clusters; this prototype’s poem did not test such a network.
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
The Test Behind Orbital AI
The test matters because it moves one part of the orbital-computing idea from proposal to hardware demonstration: a data-center-class accelerator operated in space, handled an AI prompt and returned an output. That is a useful engineering milestone for Google, but it answers only a limited question. It does not establish that space-based AI can match terrestrial systems in cost, capacity or reliability.
Google’s stated motivation is access to solar energy. Its research says solar panels in a suitable orbit can be up to eight times more productive than on Earth. That claim speaks to a real pressure on AI infrastructure: the U.S. Department of Energy estimated that data centers used about 4.4% of U.S. electricity in 2023 and projected a range of 6.7% to 12% by 2028. Those figures describe electricity demand on Earth, not a direct case for moving computing to orbit. Launch, spacecraft hardware, communications and replacement costs would also have to be counted.
Cost may determine whether the idea advances beyond experiments. Google’s paper estimates that, at current launch prices, putting one kilowatt of computing capacity in orbit would cost about $14,700 a year in launch costs. Its modeled figure falls to about $810 a year if launch prices reach $200 per kilogram, a level the paper compares with power costs for U.S. data centers. Google estimates reaching the proposed scale could require roughly 1,800 Starship launches. These are research estimates based on assumptions, not prices for an operating service or a demonstrated savings.
Popular right now
Vitamix Propel Series 750 Professional-Grade Blender, 4 Programs, Black
Dyson Car+Boat Cordless Handheld Vacuum
Breville Bambino Espresso Machine, Brushed Stainless Steel (BES450BSS)
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 described solar-powered satellites carrying TPUs and flying close enough to exchange data over laser links. Google’s initial public milestone was a learning mission with Planet involving two prototype satellites by early 2027. The October launch brought a smaller, single-satellite test earlier than that stated target, but it did not demonstrate the planned interconnected system.
Google published a peer-reviewed version of its research in Joule on October 1, the date it said the prototype reached orbit. The company’s design sketches clusters of 81 satellites at about 650 kilometers altitude, with neighboring spacecraft roughly 100 to 200 meters apart. Google says a bench test of a pair of optical transceivers reached 1.6 terabits per second. That is a ground test of communications hardware, not evidence that an 81-satellite cluster is operating in space.
Other companies have also tested computing hardware in orbit, so the broader idea predates this week’s result. The distinction Google emphasizes is its proposed use of tightly linked satellites to work as a cluster rather than running AI workloads on one spacecraft at a time. Whether that architecture can function reliably in orbit remains unproven.
““The first-ever AI response from a data center class TPU in orbit.””
— Google, describing the October test
Costs and Hardware Risks Remain
Several technical and economic questions remain open. The current satellite’s power and heat limits permit only short runs on one TPU at a time. Google has not shown that the hardware can sustain larger workloads over long periods, or that its planned laser-linked cluster can coordinate computing jobs while satellites move in orbit. The proposed system must also communicate with ground stations and deliver useful data within practical limits.
Heat management in a vacuum, radiation damage, repairs and replacement are unresolved challenges identified in the source material. Space debris is another operating concern: the material cites about 47,000 tracked objects in orbit, without specifying the date or categories included in that count. The launch-cost estimates depend on much cheaper future launches and large-scale deployment; the stated assumptions have not been demonstrated by an operational orbital data center.
It is also not yet clear how the satellite performed across all initial tests, how frequently it can run its TPU, or whether the returned poem required any processing outside the spacecraft. Google has not released enough detail to independently evaluate those points. The haiku confirms a limited demonstration, not a full account of the system’s capabilities.
Thermal Tests and Twin Satellites
Google said the next work will test TPU performance under thermal stress and radiation exposure. If those tests proceed as planned, the 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, addressing capabilities that the single-satellite demonstration did not cover.
Those results should offer a clearer measure of whether Google can keep the hardware operating in orbit and exchange data between spacecraft. The company has not announced a commercial deployment schedule or shown that the estimated launch economics can be achieved. For now, Project Suncatcher remains a research program: the haiku is evidence of an initial in-orbit AI response, while the larger questions about scale, cost and reliability await further tests.
Key Questions
What did Google’s satellite do?
Google said it used a TPU running Gemma 4 to generate a haiku and transmit the response to a ground station during a pass over Svalbard. The company described it as an initial system test.
Was this the first AI model run in space?
No. Source reports say Starcloud launched a satellite carrying an Nvidia H100 in November 2025 and later ran a small language model and Google’s Gemma model in orbit. Google’s narrower claim concerns an AI response from its data-center-class TPU.
Is Google operating a data center in orbit?
No orbital data center has been demonstrated by this test. The satellite is a prototype with limited power, and Google’s larger plan would require multiple satellites linked together.
What is Google testing next?
Google said it plans to test how TPUs handle heat and radiation in orbit. Its next stated milestone is a pair of interconnected satellites in 2027, intended to test higher power and optical links.
Source: ThorstenMeyerAI.com
