🔍 Read the full analysis: How One Google Satellite Haiku Opens Up Questions About AI In Space on ThorstenMeyerAI.com
Get privacy and security gear 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 its Gemma 4 model during a pass over Svalbard. The test shows that a small AI computing payload can operate in orbit, but it does not establish that space-based data centers are practical or economical.
Google says a small satellite carrying one of its TPU chips generated a haiku in orbit this week, marking the first reported response from its Project Suncatcher prototype. The test, using Google’s Gemma 4 model during a ground-station pass over Svalbard, shows that the company’s hardware can run an AI task in space; it does not show that an orbital data center is ready to operate at commercial scale.
The satellite launched on October 1 aboard SpaceX’s Transporter-18 rideshare mission. Google developed the prototype with satellite company Planet. Following several weeks of commissioning, engineers sent a prompt asking Gemma 4 to write a haiku about space. Google said the satellite returned the poem as it passed over Svalbard, Norway, and that engineers also tested space-themed limericks and asked the payload to report its technical status.
The poem was: “Blue marble below / Data streams in silent grace / Starlight feeds my core.” Google described the response as the mission’s “Hello World” moment and called it “the first-ever AI response from a data center class TPU in orbit.” That wording refers specifically to a response from a TPU, not to the first AI computation in space.
TechCrunch reported that the satellite carries four TPUs and generates up to about one kilowatt of power. Power and heat limits mean it can operate one chip at a time, in short bursts. Google said the satellite was operating as expected. The company’s next tests are intended to evaluate how the chips handle radiation and thermal stress in orbit.
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 Economics Behind Orbital AI
The experiment matters as a test of whether AI computing might eventually use solar power collected in orbit, rather than adding all of its energy demand to terrestrial electricity systems. Google argues that solar panels in a suitable orbit could be up to eight times more productive than on Earth. That is the company’s rationale for investigating space-based computing, not a result established by this small prototype.
The financial hurdle is substantial. In Google’s research, the estimated annual launch cost for one kilowatt of orbital capacity is about $14,700 at current prices. The paper estimates that figure could fall to about $810 if launch costs reach $200 per kilogram, a level Google compares with the cost of electricity for US data centers. The lower-cost scenario depends on future launch prices; it is not the current cost of operating the satellite.
Google’s research says reaching the proposed scale would require roughly 1,800 Starship launches. The haiku test therefore demonstrates a limited technical capability, while leaving open the central commercial question: whether launching, operating and replacing orbital computing hardware can become affordable enough to compete with facilities on the ground.
Products Worth Considering
Amazon Kindle Scribe Colorsoft 64GB (1st Generation)
Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Suncatcher to a Single Satellite
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. The proposal envisioned solar-powered satellites carrying TPUs, flying close together and using laser links to exchange data as a cluster. Google initially described a “learning mission” with Planet involving two prototype satellites, with launch planned by early 2027. The October launch brought forward a smaller, single-satellite test.
Google published a peer-reviewed version of its research in the journal Joule on October 1, according to the supplied source material. The company’s longer-term design is much more ambitious than the current payload: its paper sketches groups of 81 satellites at roughly 650 kilometers altitude, with neighboring craft separated by 100 to 200 meters. Google says a bench test of a pair of optical transceivers reached 1.6 terabits per second; that is a ground test, not a demonstrated link between operational satellites.
The experiment is also not the first reported AI computing activity 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. The distinction in Google’s claim is the use of a data-center-class TPU and its proposed approach of linking satellites into a computing cluster.
““the first-ever AI response from a data center class TPU in orbit.””
— Google, describing the prototype’s response
The Limits Still Untested
The current demonstration does not establish how long the TPU can operate reliably, how it would be repaired, or how a larger constellation would handle heat, radiation and communications. A satellite can radiate heat differently from a terrestrial server, and the source material identifies thermal management as an unresolved engineering challenge. Google’s upcoming tests may provide information about hardware performance, but no results from those tests have been reported yet.
Economics are also uncertain. The launch-cost estimates depend on prices that may change, while the proposed cluster would require many launches and reliable links among satellites and to ground stations. Space debris adds another concern: the source material cites about 47,000 tracked objects in orbit, but does not specify the tracking source or date for that count. It remains unclear whether the prototype’s short operating bursts can be extended, or whether the proposed system can match the cost and performance of Earth-based data centers.
Tests Before Satellite Clusters
Google says it will continue testing the prototype, with a focus on how its TPUs respond to radiation exposure and thermal stress. The company’s next stated milestone is a pair of interconnected satellites in 2027. TechCrunch reported that the twin-satellite mission is expected to generate more power and test optical links between spacecraft, a step toward the cluster design in Google’s research.
Those tests will be more informative than the initial poem about whether satellites can exchange data and run workloads together in orbit. Google has not established a commercial deployment schedule or confirmed that the cost estimates needed for a large system can be reached. For now, the haiku is evidence of a working, small-scale test—not evidence that data centers are moving into space.
Key Questions
What did Google’s satellite do?
Google says a satellite carrying a TPU chip ran its Gemma 4 model and returned a haiku after engineers sent it a prompt during a pass over Svalbard.
Was this the first AI computation in space?
No. The supplied source cites reports that Starcloud launched a satellite with an Nvidia H100 GPU in 2025 and later ran Google’s Gemma model in orbit. Google’s narrower claim concerns an AI response from a data-center-class TPU.
Can the satellite run all four TPUs at once?
No. TechCrunch reported that the craft can produce up to about one kilowatt and can operate one TPU at a time, in short bursts, because of power and heat constraints.
Does the haiku prove orbital data centers are viable?
No. It shows that a small AI payload generated a response in orbit. The cost, reliability, heat management, radiation exposure, repair options and satellite-to-satellite communications needed for a larger system remain open questions.
What is Google testing next?
Google says it plans further tests of chip performance under radiation and thermal stress. Its next stated milestone is a two-satellite mission in 2027 to test interconnected spacecraft, according to the supplied material.
Source: ThorstenMeyerAI.com
