[Digital Today intern reporter Seung-a Yoo] Google will launch a satellite carrying its in-house artificial intelligence (AI) computing chips into low-Earth orbit on Oct. 1. The first orbital test will verify whether the AI chips can operate in space. It is the first step in 'Project Suncatcher,' which in the long term aims to build AI data centres in space.

On Sept. 24 (local time), Google and foreign media reports said Google plans to launch the AI computing satellite on Oct. 1 aboard SpaceX's Falcon 9 rocket, as part of the Transporter-18 rideshare mission. The satellite, co-developed with Planet, carries 4 of Google's Tensor Processing Units (TPUs). The 4 TPUs are known to have computing power equivalent to 1 data centre server.

The satellite will run on about 1 kilowatt (kW) of solar power and handle simple AI queries. The test operation period is planned at about 1 year. The satellite itself is expected to remain in orbit for up to 6 years before re-entering the atmosphere.

The key aim of the mission is to see whether the TPUs can withstand the space environment. During the rocket's insertion into low-Earth orbit, the spacecraft can be subjected to acceleration of up to about 10G. Individual components can receive forces of 50 to 100G. Google previously ran a test that shook the satellite along three axes to simulate launch vibrations, and said the hardware withstood it.

Radiation is also a key test target. Google irradiated the TPUs with a proton beam at the Crocker Nuclear Laboratory at the University of California, Davis to simulate the space radiation environment. Radiation can cause errors such as 'bit flips,' where 0s and 1s in semiconductor data are reversed. Google said initial tests showed Trillium TPUs withstood a total radiation dose exceeding the level expected over a 5-year space mission. Whether they can operate over the long term in real space conditions will be further verified through this orbital test.

Cooling is another challenge. Because space is a vacuum, it is difficult to expel heat using airflow as in ground-based data centres. Google is developing a cooling system that combines heat pipes and radiators and plans to test its performance in space. The TPUs on the satellite are known to need to stop operating after about 15 minutes to cool down. As a result, the mission is an experiment to confirm the technical conditions for operating AI computing hardware in space, rather than a step to run a large-scale AI data centre.

Project Suncatcher was first unveiled in November last year. Google is studying ways to build infrastructure that mounts TPUs on solar-powered satellites, links multiple satellites and performs machine-learning computations in space. Google said low-Earth orbit satellites can receive sunlight almost continuously, and in certain orbits power productivity can be up to 8 times higher than ground-based solar panels.

Google is also developing inter-satellite communications technology. High-speed data communications between satellites are needed for multiple satellites to operate as a cluster and perform AI computing. Google is researching laser-based free-space optical communications. It plans to launch 2 additional satellites in 2027 to test related technologies.

Competition is also continuing over space-based AI data centres. SpaceX in February presented a plan to build space-based AI infrastructure as it announced integration with AI company xAI. SpaceX is known to have applied to the U.S. Federal Communications Commission (FCC) for a plan to deploy up to 1,000,000 satellites for AI computing. Blue Origin also applied to the FCC in March for 'Project Sunrise,' a plan to deploy about 52,000 satellites capable of AI computing.

The space data centre concept is drawing attention as power demand for AI computing rises. According to Google's 2025 research, if launch costs continue to fall through the mid-2030s to reach below $200 per kilogram, the launch and operating costs of space data centres could approach a level similar to the electricity costs of ground-based data centres. Technical challenges remain, including thermal management, high-speed communications between satellites and the reliability of in-orbit systems.

This satellite launch is an early step to test the feasibility of Project Suncatcher. Google plans to confirm TPU durability, radiation effects and cooling performance in real space conditions, and then reflect the results in future designs for satellite-based AI computing infrastructure.