🔍 Read the full analysis: Google’s Data Center Satellite And The Reality Of AI In Space on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Google says Gemma 4 generated a haiku using a data-center-class TPU aboard a small satellite launched on October 1, 2026. The test shows that an AI chip can run in orbit, but the satellite’s limited power and unresolved costs, cooling, radiation and repair challenges leave large-scale space computing far from proven.
Google says its Gemma 4 model generated a haiku aboard a satellite carrying one of the company’s data-center-class TPU chips, a small but real test of its plan to explore AI computing in orbit. The satellite launched on October 1, 2026, and the response this week marks an early milestone for Project Suncatcher—not a demonstration of a working orbital data center.
Google described the prompt-and-response test as its “first-ever AI response from a data center class TPU in orbit.” During a pass over Svalbard, Norway, engineers sent a request for a space-themed haiku to the satellite and received the generated lines at a ground station. Google said it also tested limericks and asked the payload to report technical status. The company’s account was published October 30; a Google spokesperson told TechCrunch that the satellite was “operating nominally.”
The spacecraft was built with satellite company Planet and flew on a SpaceX Transporter-18 rideshare mission. According to TechCrunch, it carries four TPUs but can produce up to about one kilowatt of power, limiting operations to one chip at a time. The report described chip runs of roughly 15 minutes before launch and about 20 minutes during the recent test. Google said the satellite spent about three weeks in commissioning, including checks of its host computer, TPU trays and early machine-learning tests.
The test is a step beyond simply placing hardware in orbit, but it does not show that a cluster of satellites can perform sustained, large-scale AI workloads. Google says the next tests will examine how its TPUs handle thermal stress and radiation. Its planned follow-up is a pair of interconnected satellites in 2027, which would test links between spacecraft as well as operation with more power. Those milestones have not yet established a commercial system.
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 Chip Test, Not an Orbital Data Center
The experiment matters because it tests whether a chip designed for data-center workloads can operate in the space environment and communicate with engineers on Earth. Google’s broader proposal is to use solar-powered satellites linked by lasers as a distributed computing system. If that approach could eventually be made reliable and economical, it might offer another way to supply computing capacity as demand for AI grows.
But the prototype’s scale shows the distance between that idea and current capability. A satellite producing about one kilowatt cannot match the power available at a terrestrial data center, and brief chip runs are not evidence of continuous service or large-scale model training. The value of the test is technical learning: it can help identify operational limits before Google attempts more complex linked systems.
Cost is another major constraint. In its research, Google estimates that, at current launch prices, the annual launch cost associated with a kilowatt of orbital capacity would be about $14,700. The paper estimates about $810 per kilowatt-year if launch costs fall to $200 per kilogram. Those figures are modelled estimates, not operating costs measured from this prototype. The lower estimate depends on a substantial fall in launch costs and, according to the source material, a build-out requiring roughly 1,800 Starship launches.
Products Worth Considering
HP 4×6 Direct Thermal Shipping Label Printer, 203 DPI
Apple Mac mini Desktop Computer with M4 chip, 10‑core CPU and GPU, 24GB Unified Memory, 512GB SSD, Gigabit Ethernet
FLEXISPOT 42-Inch Electric Standing Desk Converter
As an affiliate, we earn on qualifying purchases.
From Suncatcher to the First Prompt
Google introduced Project Suncatcher in a research post and preprint on November 4, 2025. The proposal described satellites carrying TPUs, flying in close formation and using laser links to share data. Google then identified a learning mission with Planet and two prototype satellites, with a target of early 2027. The first prototype reached orbit earlier, on October 1, 2026, but the initial test was narrower than the proposed network: one satellite responding to prompts relayed through a ground station.
Google’s research argues that solar panels in a suitable orbit could be up to eight times more productive than on Earth. The motivation is the energy needed by computing facilities: the source material cites U.S. data centers as using about 4.4% of national electricity in 2023, with the Department of Energy projecting a range of 6.7% to 12% by 2028. These estimates describe the broader U.S. electricity picture; they do not show that orbital computing would reduce energy use or costs in practice.
Google is not the first company to put AI-capable hardware 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. Google’s narrower distinction is its claim about a response from a data-center-class TPU. Its longer-term proposal also emphasizes tightly spaced satellite clusters and high-speed optical links, rather than treating each spacecraft as a standalone computer.
““the first-ever AI response from a data center class TPU in orbit.””
Power, Cooling and Cost Remain Open
The test does not settle whether orbital AI computing can scale or compete with terrestrial data centers. The satellite’s power and operating limits are far below those implied by Google’s proposed multi-satellite clusters. The source material also identifies heat management in vacuum, radiation damage, lack of repair access and reliable links to ground stations as unresolved challenges. Google’s planned thermal and radiation tests may provide early evidence, but their results have not yet been reported.
Google’s cost estimates depend heavily on future launch prices, and the lower-cost scenario is not the present-day economics of the prototype. It is also not clear from the available material how a large constellation would handle maintenance, replacement, service interruptions or the risks posed by orbital debris. The source cites about 47,000 tracked objects in orbit, but does not specify a time window or comparison basis for that count.
The first response confirms a specific event: Gemma 4 generated text on a TPU in orbit and the result reached the ground. It does not independently establish the performance, reliability or economics of a sustained orbital computing service. Google’s claims about the test and its future plans remain company statements; further results are needed to assess the system.
Two-Satellite Tests Planned for 2027
Google says it will continue testing the current spacecraft, including how TPUs perform under thermal stress and radiation exposure. The company’s next stated milestone is a pair of interconnected satellites in 2027. TechCrunch reported that the follow-up spacecraft are expected to generate more power and test optical links between satellites. The source material does not give a specific launch date or a detailed schedule for those tests.
The results will help show whether multiple satellites can exchange data fast enough to operate as a coordinated computing system, rather than as separate machines. Readers should look for reported test outcomes—especially sustained chip operation, link performance and power use—before treating the proposal as a practical alternative to Earth-based data centers. For now, Google has shown a brief AI response in orbit, while the larger engineering and cost questions remain open.
Key Questions
What did Google’s satellite do?
Google says it ran Gemma 4 on a TPU in orbit and returned a haiku after engineers sent a prompt through a ground station. The company also said it tested other prompts and a technical-status request.
Was this the first AI model in space?
No. The source material says Starcloud launched a satellite with an Nvidia H100 GPU in November 2025, and CNBC later reported it ran and queried Google’s Gemma model. Google’s claim is narrower: its test produced a response from a data-center-class TPU in orbit.
Is Google operating a data center in space?
No. The test used one small satellite with limited power, which TechCrunch reported restricts it to one TPU at a time and short operating bursts. A network of linked satellites remains a future research plan, not an operating data center.
What happens after this test?
Google says it will test the current satellite’s response to thermal stress and radiation. Its next stated milestone is two interconnected satellites in 2027, intended to test higher power and links between spacecraft.
Source: ThorstenMeyerAI.com
