GUANGZHOU, Jan. 27 (Xinhua) — A Chinese private aerospace company has successfully introduced a general-purpose AI model to an orbiting satellite, marking a major milestone for space-based computing.
Guoxing Aerospace Technology announced at a seminar on Monday that it has uplinked Alibaba’s Qwen3 large-scale language model to its first space-based computing center, enabling end-to-end inference tasks entirely in orbit.
“This is the world’s first example of deploying a general-purpose large-scale AI model from ground control to operational satellite constellations in orbit,” said Wang Yabo, executive vice president of the Chengdu-based startup.
Last May, China launched into orbit a new constellation of 12 space computing satellites, the first cluster of Guoxing Aerospace’s space computing project.
During testing, the Qwen3 model completed multiple experiments in which questions were transmitted from Earth to the satellite, processed onboard, and results returned to the ground station. All of this was completed within just 2 minutes.
As AI fuels an insatiable appetite for computing power, intelligent computing power is being pushed into space, creating a new field in the technological race. In November, a SpaceX rocket orbited the Starcloud-1 satellite powered by Nvidia GPUs.
Wang outlined the company’s ambitions to build a vast network of 2,800 dedicated computing satellites by 2035.
The planned satellite constellation includes 2,400 inference satellites and 400 training satellites to be placed in sun-synchronous, dawn-to-dusk, and low-inclination orbits at altitudes of 500 to 1,000 kilometers.
The constellation is designed to employ laser intersatellite links to facilitate high-speed data transfer and aims to deliver 100,000 petaflops of inference computing and 1 million petaflops of training compute worldwide.
Wang said the second and third satellite clusters will be deployed this year, with the goal of completing a 1,000-satellite network by 2030. ■
