Guoxing Yuhang launches eight space computing satellites to test on-orbit AI model training

Guoxing Yuhang successfully launched the second group of eight satellites under its “XingSuan” space computing plan at 19:40 Beijing time on October 14, 2026. The satellites were carried by a Zhuque-2 Improved Y4 rocket from the Dongfeng Commercial Space Innovation Pilot Zone and entered their intended orbits.

All eight satellites were developed by the company and are equipped with high-performance computing payloads, high-speed inter-satellite and satellite-ground laser communication payloads, and low-light cameras. They will be used to verify technologies including 10,000-watt-level liquid cooling fluid loops, large deployable radiation cooling wings, a global laser communication network, and on-orbit training of foundation models.

The XingSuan plan is co-built with 108 universities, research institutes, listed companies and investment institutions. It aims to deploy 2,800 computing satellites, of which 2,400 are planned for inference computing and 400 for training computing, to form a space-based training and inference computing network. The company plans to complete a commercial constellation of more than 1,000 satellites before 2030 and finish the full network by 2035, providing space computing resources for AI model inference and training.