GuoXing Aerospace Technology has achieved a groundbreaking milestone in space-based computing by deploying a general-purpose AI model aboard its satellite constellation, the company announced this week. The Chengdu-based firm successfully integrated Alibaba's Qwen3 large language model into its inaugural space computing center, enabling real-time reasoning tasks entirely in orbit.
Executive Vice President Wang Yabo revealed that the system processed queries transmitted from Earth and returned results within two minutes during recent trials. This development positions China at the forefront of an emerging global race to expand computing infrastructure beyond terrestrial limits.
The project builds on last May's launch of 12 specialized satellites, with plans to deploy 1,000 units by 2030 and a full network of 2,800 satellites by 2035. The proposed constellation will combine laser-linked inference and training satellites across multiple orbital planes, aiming to deliver unprecedented computational power equivalent to 100,000 petaflops for AI inference tasks.
This advancement comes as global interest in space-based computing intensifies, following SpaceX's November 2025 launch of GPU-equipped satellites. Analysts suggest such systems could revolutionize fields from climate modeling to financial forecasting by leveraging the unique advantages of orbital infrastructure.
