Project Suncatcher, Google’s first satellite data center, has created a work of poetry from space, according to reports on Friday.
The technology behind this achievement is DeepMind’s Gemma 4. This confirms that artificial intelligence is functioning in space. However, there remain questions about its economic viability.
One TPU, 20 minutes, and three lines of verse
Google launched the prototype on October 1. It flew aboard a SpaceX rideshare, with the work executed in partnership with Planet. Approximately three weeks later, after a thorough check, engineers asked Gemma 4 to create a description of space. It responded:
“Blue marble below
Data streams in silent grace
Starlight feeds my core”
The satellite carries four TPUs, but there’s a catch: it can run only one at a time, and for about 20 minutes. It generates up to 1 kilowatt of electricity, TechCrunch reported . Engineers also asked the model to write limericks and report on conditions in orbit. It’s a promising first test. However, running a commercial data center in space is much more than about writing poetry.
It makes sense that Google has turned its sights to the sky. According to a study published in Nature , electricity consumption by AI-dedicated data centers in 2024 was at 118 terawatt hours and could soon exceed 239-295 terawatt hours by 2030.
Goldman Sachs estimates a growth of 170% in electricity needed for data centers around the globe between 2025 and 2030. JLL notes that connecting to grids can take anywhere from 2 to 10 years. Orbit is able to provide solar power without waiting times. The main challenge is getting the equipment there and keeping it running.
Which workloads fit, and who gains
But space is not ideally suited to all forms of computing. JLL says that its best candidates are AI training, batch processing, and processing of the data generated in space. For real-time transactions, it is essential to be close to the users.
This narrows down the market, given that costs are still high. BCG estimates the orbital cost premium today to lie between 2.5 and 3 times higher. But it estimates that space could eventually claim 10 to 15 percent of the market for AI data centers by 2040. That’s a forecast, not an absolute certainty.
For space-based data centers to capture that projected market share, operators must overcome three major hurdles: high costs, the effects of heat and radiation on chips, and the difficulty of transferring data between satellites. Google’s haiku proves an AI model can run in orbit. But turning that early success into a commercial business is another challenge.
Heat, radiation, and the physics that won’t cooperate
Before orbital computing can become a commercial business, operators must overcome some serious technical challenges. Cooling is one of them. With no air to carry heat away, servers need large radiators. Brookings cites an estimate of 2.15 million square feet of radiator area for an orbital data center.
But cooling isn’t the only concern. Ultraviolet exposure and space debris threaten equipment, while repairs are difficult. Radiation can also cause memory “bit flips” and corrupt results, TechCrunch noted. These risks highlight the need for radiation-hardened chips, as discussed by Semiconductor Engineering . Even the environmental benefits need scrutiny. An ACM study examines the carbon trade-offs of computing in orbit.
Google Project Suncatcher: Orbital AI Timeline, Costs and Risks
SpaceX, Starcloud and the 2027 twin-satellite test
Despite those challenges, companies are pushing ahead. Cryptopolitan reported SpaceX is seeking about $40 billion for Nvidia chips for AI data centers on Earth and potentially in orbit. Financing remains proposed. Starcloud has flown an H100 into orbit. Cowboy Space is pursuing single-satellite inference, TechCrunch reported.
Google’s plan is different: close-flying satellites sharing work through lasers. If current tests succeed, it plans two higher-power spacecraft in 2027 to test those links. The next breakthrough won’t be another poem. It will be proof that the satellites can work together.
If you’re reading this, you’re already ahead. Stay there with our newsletter .
