🔍 Read the full analysis: Google’s First Data Center Satellite Wrote A Haiku. Here’s The AI Context 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 with the Gemma 4 AI model during a test above Svalbard. The demonstration shows that a data-center-class TPU can run in orbit, but the prototype’s limited power and computing time are far from a working orbital data center.
Google says a satellite carrying one of its data-center-class TPU chips has generated an AI-written haiku in orbit, marking the first reported computing result from its Project Suncatcher prototype. The test, using Google’s Gemma 4 model as the satellite passed over Svalbard, shows the hardware can run an AI task in space; it does not demonstrate a large or commercially useful orbital data center.
The satellite sent down the three-line poem: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” Google described the moment as its “Hello World” test and called it the “first-ever AI response from a data center class TPU in orbit.” The company said engineers also tested space-themed limericks and asked the system to report its technical status. A Google spokesperson told TechCrunch the satellite was “operating nominally.”
The prototype was built with satellite company Planet and launched on October 1 aboard SpaceX’s Transporter-18 rideshare mission. After roughly three weeks of commissioning, Google said it powered up the host computer and TPU hardware and began initial machine-learning tests. The satellite carries four TPUs, but TechCrunch reported that its power supply produces up to about 1 kilowatt, limiting operation to one chip at a time. Test runs have lasted around 15 to 20 minutes.
This is a demonstration at small scale, not a functioning network of space-based computers. Google’s longer-term plan involves satellites carrying TPUs and communicating through laser links, so multiple spacecraft could work together. The company’s next stated step is to test two interconnected satellites, currently planned for 2027.
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 one part of Google’s space-computing proposal from research into an on-orbit hardware trial: a TPU ran an AI model and returned a result through a ground station. That gives Google a way to begin studying how its chips behave under radiation and thermal stress, conditions that differ sharply from those in a terrestrial data center.
The potential motivation is electricity. Google’s research argues that solar panels in a suitable orbit could generate more power than comparable panels on Earth. The company is exploring whether that advantage could help meet rising demand from AI computing. But the haiku does not establish that orbit-based computing is cheaper, more reliable, or ready to ease pressure on power grids. The prototype’s approximately 1-kilowatt capacity is tiny beside the power needs of large data centers, and it operates only in short bursts.
Google’s own cost estimates underline the gap. Its research puts the annual launch cost of delivering a kilowatt of capacity to orbit at about $14,700 at current prices. In a lower-cost scenario of $200 per kilogram for launch, the paper estimates about $810 a year per kilowatt, near the price of electricity for U.S. data centers. That lower estimate depends on a major fall in launch costs; the paper says reaching the scenario would require roughly 1,800 Starship launches. These are modeled estimates, not demonstrated operating costs.
Products Worth Considering
Bose QuietComfort Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones, Black
DJI Mini 5 Pro Fly More Combo Plus with DJI RC 2
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
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, describing a future constellation of solar-powered satellites equipped with TPUs and connected by optical links. Its initial stated milestone was a learning mission with Planet involving two prototype satellites by early 2027. The first prototype launched earlier, on October 1, 2026, and Google published a peer-reviewed version of its research in the journal Joule on the same date.
Google’s claim about being first is specific to a response from a data-center-class TPU. It is not a claim that no AI chip or model has operated in space. 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. The distinction in Google’s project is its proposed use of closely spaced satellites linked to operate as a larger computing system, rather than AI computing on a single spacecraft.
Google’s research describes a possible cluster of 81 satellites flying within about a kilometer of one another at an altitude near 650 kilometers. It also reports a bench test of a laser-link pair reaching 1.6 terabits per second. That laboratory result is not evidence that a satellite constellation has already achieved the same performance in orbit. The current prototype is a first hardware step toward testing the concept, not a demonstration of the proposed cluster.
““first-ever AI response from a data center class TPU in orbit””
The Engineering Gaps Ahead
Google has not shown that multiple satellites can yet operate as a coordinated AI cluster in orbit. The prototype’s first response confirms a limited task, but its results do not establish sustained computing performance, reliability over long periods, or cost competitiveness with Earth-based facilities. Google has also not reported how the TPU performs across extended exposure to radiation and temperature extremes.
Other challenges include removing heat in a vacuum, repairing or replacing equipment that fails, maintaining dependable links with the ground, and coordinating close-flying satellites. The source material cites about 47,000 tracked objects in orbit, a figure that makes the orbital environment relevant to the project’s safety and operations, but does not provide a time-specific count or clarify the tracking threshold. It is also unclear how much power the next prototypes will produce, how often they will communicate, or what their operating costs would be. Google’s projected launch-cost savings depend on future launch prices that have not been established.
Tests Before the Twin-Satellite Trial
Google says it will continue testing the prototype, including how the TPU responds to radiation and thermal conditions in orbit. The next planned milestone is a pair of interconnected satellites in 2027. According to the source material, those spacecraft are expected to generate more power and test optical links between satellites, addressing two requirements for Google’s broader concept.
The results will help show whether the system can move beyond isolated computing tasks toward coordinated work between spacecraft. Until Google reports the outcome of those tests—and provides evidence on performance, durability, power and cost—the haiku remains a successful early demonstration, not proof that data centers are moving into space at commercial scale.
Key Questions
What did Google’s satellite do?
Google says its satellite used a TPU chip running the Gemma 4 model to generate a haiku, which was then sent to a ground station as the spacecraft passed over Svalbard.
Was this the first AI system in space?
No. The claim is narrower: Google calls it the first AI response from a data-center-class TPU in orbit. The source material says Starcloud previously launched an Nvidia H100 and ran AI models in space.
How much computing power does the prototype have?
The satellite carries four TPUs, but TechCrunch reported that its power supply reaches about 1 kilowatt, so it can run only one chip at a time in short bursts of roughly 15 to 20 minutes.
What is Project Suncatcher trying to achieve?
Google is studying whether solar-powered satellites carrying TPUs could be linked by lasers and work together as an orbital computing system. That larger system has not yet been demonstrated in space.
What happens after the haiku test?
Google says it will test the prototype under radiation and thermal stress. The next stated milestone is a two-satellite test planned for 2027, including trials of inter-satellite optical links.
Source: ThorstenMeyerAI.com
