🔍 Read the full analysis: Google’s Satellite Haiku: A Window Into AI In Space, Not The Whole Picture on ThorstenMeyerAI.com
Get everyday essentials 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 its Project Suncatcher prototype has returned a response from Gemma 4 running on a TPU in orbit. The test shows that a small AI workload can run in space, but it does not establish that orbital data centers are practical: power, cooling, launch costs, radiation and satellite links remain major challenges.
Google says Gemma 4 generated a haiku aboard a satellite carrying a Google TPU, a small but concrete test of Project Suncatcher, the company’s effort to study whether AI computing could operate in space. The response was sent to a ground station during a pass over Svalbard, Norway; it demonstrates an in-orbit AI workload, not a functioning orbital data center at commercial scale.
The prototype, built with satellite company Planet, launched October 1 on SpaceX’s Transporter-18 rideshare mission, according to the source report. Google said the satellite made contact and was operating as expected. After several weeks of commissioning checks, engineers sent a prompt asking Gemma 4 to write a haiku about space. Google described the returned poem as its “Hello World” moment for the mission.
According to TechCrunch’s account of the hardware, the satellite carries four TPUs but produces up to about one kilowatt of power. It can operate only one chip at a time, in bursts of roughly 20 minutes, reflecting limits on available power and heat management. Google said engineers also tested space-themed limericks and asked the payload to report technical status. A Google spokesperson told TechCrunch that the satellite was “operating nominally.”
Google’s next planned work is to test how the TPUs perform under thermal stress and radiation exposure. The company has described a subsequent milestone involving two interconnected satellites in 2027. Those spacecraft are intended to test optical links between satellites, a step toward the clustered computing architecture in Google’s proposal.
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 Tests the Orbital Hardware
The result matters as an engineering demonstration, not as proof that AI data centers are ready to move off Earth. The satellite appears to have completed a simple prompt-and-response task using an AI model and a data-center-class TPU in orbit. That gives Google a real environment in which to examine computing hardware, communications and operating conditions beyond laboratory tests.
The economic case is unresolved. Google’s research estimates that, at current launch prices, launch costs alone amount to about $14,700 per kilowatt per year for orbital capacity. The paper estimates that this could fall to about $810 if launch costs reach $200 per kilogram. These are modeled figures, not demonstrated operating costs, and the lower estimate depends on a substantial fall in launch prices. The source says reaching the proposed scale would require roughly 1,800 Starship launches.
For readers watching AI’s energy demand, the project points to one proposed way of accessing solar power beyond the limits of terrestrial sites. It does not yet show that space-based computing can compete on cost or reliability. The prototype’s limited power and short operating bursts underline how far a successful demonstration remains from a practical service.
Products Worth Considering
Anker SOLIX C300 Portable Power Station, 288Wh
Jackery Explorer 500 v2 Portable Power Station, 512Wh, 500W
Jackery Explorer 2000 v2 Portable Power Station, 2042Wh
As an affiliate, we earn on qualifying purchases.
From Suncatcher Proposal to Flight
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. Its concept is a constellation of solar-powered satellites carrying TPUs and connected by laser links, with nearby spacecraft working together like a computing cluster. Google’s research sketches groups of 81 satellites flying within about one kilometer of one another at an altitude of roughly 650 kilometers. The company says a bench test of a pair of optical transceivers reached 1.6 terabits per second; that is a laboratory result, not a reported in-orbit link performance.
The flight came earlier, and on a smaller scale, than the previously described milestone. Google had said it planned a learning mission with Planet involving two prototype satellites by early 2027. Instead, a single prototype launched in October 2026. The source report says the satellite flew alongside more than 100 other spacecraft on the rideshare mission.
Google’s claim should also be read narrowly. The source material reports that Starcloud launched a satellite carrying an Nvidia H100 in November 2025 and later ran and queried Google’s Gemma model in orbit, according to CNBC. Google’s distinction is its claim of the first response from a data-center-class TPU in orbit, not the first AI model or AI chip in space.
““the first-ever AI response from a data center class TPU in orbit.””
Costs and Orbital Risks Remain
The demonstration does not resolve whether orbital computing can be economical or dependable at scale. Google’s cost estimates depend on launch prices, while the prototype has only about one kilowatt available and can run one TPU at a time. The source material does not report the satellite’s full mission cost, sustained computing output, or performance over a long operating period.
Other challenges cited in the source include removing heat in a vacuum, protecting electronics from radiation, repairing equipment that cannot readily be serviced, and maintaining links to ground stations and between satellites. The report also cites about 47,000 tracked objects in orbit; it does not specify a time window or counting method for that figure. The proposed close-flying clusters and their laser links have not yet been demonstrated in orbit.
Google’s haiku test also does not establish that the system can handle large-scale model training or continuous workloads. The details of the next satellite pair, including its final schedule and demonstrated power and link performance, remain subject to future testing.
Tests Before a Two-Satellite Link
Google says the immediate next step is a series of tests of TPU behavior under orbital thermal and radiation conditions. If those tests proceed as planned, the company’s next stated milestone is two interconnected satellites in 2027. TechCrunch reported that the pair is expected to generate more power and test optical links between spacecraft.
The results will help show whether the system can move beyond a single satellite performing short computing bursts. Evidence to watch for includes sustained operation, measured performance under radiation and heat, and successful high-speed communication between spacecraft. Until those tests are reported, Suncatcher remains an early research effort rather than an operational orbital data center.
Key Questions
What did Google’s satellite do?
Google says Gemma 4 generated a haiku on a satellite carrying a Google TPU. The response was downlinked during a pass over Svalbard, Norway.
Is this the first AI in space?
No. The source report says Starcloud previously launched a satellite with an Nvidia H100 and later ran Google’s Gemma model in orbit. Google’s claim is narrower: a response from a data-center-class TPU in orbit.
How powerful is the prototype?
TechCrunch reports that it has four TPUs and generates up to about one kilowatt. It can run one chip at a time, in bursts of roughly 20 minutes.
When will Google test linked satellites?
Google’s stated next milestone is a pair of interconnected satellites in 2027. The planned tests are intended to examine optical links between spacecraft; the schedule and results are not yet confirmed.
Does the test prove orbital data centers are cost-effective?
No. The haiku demonstrates a limited in-orbit computing task. Google’s cost figures are estimates, and launch expense, cooling, radiation, repairs and communications remain unresolved challenges.
Source: ThorstenMeyerAI.com
