Google will test AI chips in space for the first time: what they will test and why it’s needed

Google plans to launch a prototype satellite for the first orbital test of the Suncatcher project. The company will test whether its AI chips can operate stably in low Earth orbit conditions.

As Dilo writes, this was reported by Reuters.

What they will test

The mission is planned to take place during the SpaceX Transporter-18 launch in partnership with satellite company Planet Labs. The satellite prototype will be equipped with Google tensor processors.

In orbit, the company will test how the hardware withstands launch loads, space radiation, and extreme temperatures. Radiation can cause errors in electronics operation, including so-called bit flips when data values change.

Google has already tested its TPUs while performing AI workloads at the University of California, Davis. However, the company believes orbital testing is necessary to verify hardware operation directly in space.

Separately, Google will evaluate the chip cooling system. In vacuum it is impossible to use ordinary air cooling, so the company will combine heat pipes with radiators.

Further, Suncatcher project specialists are to explore the possibility of placing large computing power for AI in space. Among the potential benefits of this approach are access to sunlight for a significant portion of time and the lack of dependence on terrestrial energy constraints.

At the same time, experts note that the technology still faces high launch costs, engineering constraints, and issues with large-scale satellite production.

As a reminder, OpenAI collaborates with its main competitors — Anthropic and Google DeepMind — on AI safety measures amid growing concerns about economic and security threats from the technology.