Tanya Weaver Wed 5 Nov 2025
Collected at: https://eandt.theiet.org/2025/11/05/google-launch-constellation-ai-data-centres-space
Google has announced that its Project Suncatcher will explore building scalable AI data centres in space, with a prototype launch planned for 2027.
While the tech giant is busy scaling up its AI data centres on Earth, it is also eyeing up space as a future location for them.
Project Suncatcher is a ‘moonshot’ research project that aims to scale machine learning in space by launching the firm’s tensor processing unit (TPU) AI chips aboard solar-powered satellites into low-Earth orbit (LEO).
The aim is to deploy a constellation of networked satellites that will harness the sun’s power, a near-constant source of energy that emits more power than 100 trillion times humanity’s total electricity production.
According to Google, in the right orbit, a solar panel can be up to eight times more productive than on earth, and produce power almost continuously, reducing the need for batteries.
The tightly interconnected satellites in the Suncatcher constellation will be connected by high-speed free-space optical links. Google says that, in lab tests, each link has achieved 1.6Tbps of bidirectional bandwidth, which will enable the constellation to communicate and share data across multiple satellites at unprecedented speeds. Together these satellites will essentially form a space-based, high-bandwidth supercomputer in LEO.
The tech giant says it plans to launch two prototype satellites by early 2027 in partnership with satellite technology firm Planet Labs to test the hardware in orbit.
Space-based computing eliminates many of the problems associated with AI data centres including such as costs, power grid constraints and real estate limitations.
However, locating these data centres in space presents different challenges, including radiation tolerance of computer chips, thermal management and orbital mechanics.
Sundar Pichai, CEO at Google, said: “Like any moonshot, it’s going to require us to solve a lot of complex engineering challenges.
“Early research shows our trillium-generation TPUs (our tensor processing units, purpose-built for AI) survived without damage when tested in a particle accelerator to simulate low-earth orbit levels of radiation. However, significant challenges still remain, like thermal management and on-orbit system reliability.”
More testing will be needed ahead of the 2027 launch date, but Google is confident that despite the challenges we are heading towards gigawatt-scale constellations orbiting Earth.
Other tech giants such as Microsoft and Amazon are also betting on space-based computing.
However, the increasing use of satellites in LEO has been criticised by the space sector over its potential environmental impact.
The recent issue of E+T features an article that grapples with the question whether internet infrastructure should be moved into space.
Leave a Reply