🔍 Read the full analysis: Google’s Data Center Satellite Wrote A Haiku—Here’s What It Doesn’t Prove 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
Google says a satellite carrying its data-center-class TPU returned a haiku generated by the Gemma 4 model during a pass over Svalbard. The test marks an early Project Suncatcher milestone, but does not show that orbital AI computing is economical or ready to scale; power, heat, radiation, repairs and launch costs remain major constraints.
Google says a satellite carrying one of its data-center-class TPUs has returned a haiku generated by its Gemma 4 AI model, marking an early test of the company’s Project Suncatcher concept for computing in orbit. The response shows that a small AI payload can run and communicate from space; it does not establish that orbital data centers can operate at useful scale or compete economically with facilities on Earth.
The satellite launched on October 1, 2026, on a SpaceX rideshare mission, according to the source material. Google developed the prototype with satellite company Planet. After several weeks of commissioning, engineers sent a prompt during a ground-station pass as the satellite traveled over Svalbard, Norway. Google said the satellite returned the resulting three-line poem: “Blue marble below / Data streams in silent grace / Starlight feeds my core.”
The spacecraft carries four TPUs but produces up to about one kilowatt of power, according to details reported by TechCrunch. That limits operation to one chip at a time, in bursts; the source says tests have involved runs of about 15 to 20 minutes. Google described the first response as its “Hello World” moment and said the satellite was operating as expected. A Google spokesperson told TechCrunch it was “operating nominally.”
Google says engineers also tried space-themed limericks and asked the payload to report its technical status. The company’s next tests are intended to examine TPU performance under thermal stress and radiation exposure. The small-scale demonstration is a technical milestone, but it offers no direct evidence yet about sustained workloads, commercial performance or the cost of running useful AI services from 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
A Test, Not an Orbital Data Center
The result matters because it moves Google’s idea from research plans to a working, if limited, orbital test. AI computing requires substantial electricity, and demand for data-center power is a growing concern. Google’s broader proposal is to use solar-powered satellites and connect them so they can perform computing together. The haiku confirms only that a model could run on the prototype and that its output could be sent back to Earth.
The scale gap is large. The prototype’s roughly one-kilowatt power ceiling is not comparable to the demands of a terrestrial data center. The source material contrasts it with a nuclear power agreement Google signed on Earth on October 6 that adds 890 megawatts. That comparison illustrates the difference in capacity; it does not show that orbital systems could replace or meaningfully reduce the need for ground-based computing.
Cost is another unresolved test. Google’s research estimates that, at current launch prices, launch costs alone would amount to about $14,700 per kilowatt per year. In the paper’s scenario, that figure falls to about $810 if launch costs reach $200 per kilogram, a level the source describes as close to what U.S. data centers pay for power. The estimate depends on lower launch costs and does not by itself account for every expense of building, operating and replacing an orbital system.
Products Worth Considering
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
Bose TV Speaker Compact Soundbar with Bluetooth and HDMI-ARC, Black
Bose QuietComfort Wireless Noise Cancelling Headphones, Sandstone
As an affiliate, we earn on qualifying purchases.
From Suncatcher Paper to Flight
Google introduced Project Suncatcher in a research post and preprint paper on November 4, 2025. The proposal envisions solar-powered satellites carrying TPUs, flying in close formation and exchanging data through laser links. Google’s initial public milestone was a learning mission with Planet involving two prototype satellites by early 2027. The single-satellite test launched earlier, but is smaller than the linked system the project is meant to investigate.
Google’s claim should also be read narrowly. The source material says Starcloud launched a satellite carrying an Nvidia H100 GPU in November 2025, and CNBC later reported that the startup had run and queried Google’s Gemma model in orbit. Google’s distinction is that this is its first reported AI response from a data-center-class TPU in space—not the first AI model or AI chip used there.
The longer-term plan is more ambitious than the current flight. Google’s research describes clusters of satellites at roughly 650 kilometers altitude, with 81 satellites within about one kilometer and neighboring craft around 100 to 200 meters apart. The aim is to use optical links to connect them as a computing system. The paper reports a bench test reaching 1.6 terabits per second with a pair of optical transceivers; that is a ground test, not evidence that a satellite cluster has achieved that rate in orbit.
““the first-ever AI response from a data center class TPU in orbit””
— Google, describing the test
Costs and Conditions Still Untested
Google has not shown that the prototype can run sustained, high-demand workloads or that a cluster of satellites can work together reliably in orbit. The next environmental tests have yet to report results. The source material also does not provide a full cost comparison covering satellite manufacture, launch, operations, replacement and communications alongside the cost of terrestrial computing.
Several engineering challenges remain open: removing heat in a vacuum, protecting hardware from radiation, maintaining links to Earth and dealing with failures that may be difficult or impossible to repair. The source material cites about 47,000 tracked objects in orbit, but gives no date or definition for that count; it should be treated as a reported figure, not a measure of collision probability. The extent to which orbital congestion would affect this specific design is not established.
Twin-Satellite Link Test Planned
Google says the current satellite will undergo further tests of TPU performance under thermal stress and radiation. If those tests go well, the next stated milestone is a pair of interconnected satellites in 2027. According to TechCrunch, the pair is expected to generate more power and test optical links between spacecraft. Those results would address whether two satellites can communicate and compute together, but would still fall well short of proving that a large orbital data center is practical.
The key developments to watch are whether the hardware remains reliable through longer operations, whether the optical links perform as intended in orbit, and whether launch costs fall enough to change Google’s economic estimates. Until those questions are answered, the returned poem is best understood as a successful small-scale demonstration—not proof that AI data centers are moving to space.
Key Questions
What did Google’s satellite do?
Google says it ran its Gemma 4 model on a TPU aboard the satellite and returned a haiku during a ground-station pass over Svalbard.
Was this the first AI model in orbit?
No. The source material says Starcloud launched a satellite carrying an Nvidia H100 in 2025 and later ran Google’s Gemma model in orbit. Google’s narrower claim is that its satellite produced the first response from a data-center-class TPU in orbit.
Does the test show that orbital data centers are commercially viable?
No. It shows that a small payload could run a model and return its output. It does not establish long-term reliability, useful computing capacity at scale or competitive costs.
What happens next?
Google says it will test the current satellite under thermal and radiation stress. The company’s next stated milestone is a two-satellite connected test in 2027, subject to the results of the current work.
Source: ThorstenMeyerAI.com
