On October 14, 2026, at 19:40, the Zhuque-2 Improved Rocket Y4 lifted off from the Dongfeng Commercial Space Innovation Test Zone, successfully placing 10 satellites into their designated orbits. Among them, eight are computing satellites from Guoxing Aerospace’s “XingSuan” Program Group 02, and two are Blue Tower Intelligence optical‑communication test satellites. This launch was not merely about increasing the number of satellites in orbit; it marked China’s first large‑scale validation of the “space‑on‑space computation” model—where satellites perform AI computations onboard and transmit only decision results, rather than massive volumes of raw data.
The “XingSuan” Group 02 satellites carry high-performance computing payloads, high-speed inter-satellite and satellite-to-ground laser communication payloads, and low-light cameras, with a focus on validating four key technologies: a ten-kilowatt-class liquid-cooling fluid loop, ultra-large deployable radiative cooling wings, an all‑domain laser communication network, and on-orbit training of foundational large models. Also launched into orbit simultaneously are the first two compute‑power satellites of the Pazhou Space Intelligent Computing Constellation; this constellation is planned to comprise 24 satellites and is slated to become operational within three years. With this launch, the “XingSuan” constellation now boasts a total on-orbit computing power of 50P (500 quadrillion operations per second), effectively establishing a preliminary space‑based computing node capable of supporting real-time decision‑making.
Multiple technological approaches are advancing in parallel: Zhejiang Laboratory’s “Three-Body Computing Constellation” launched its first 12 satellites in May 2025. After nine months of on-orbit testing, it has achieved four core capabilities—network formation, computing, model deployment, and scientific payload validation—and has completed the deployment of 10 AI models. Meanwhile, the Space Server developed by Beijing University of Posts and Telecommunications’ TianSuan Constellation team has been running stably on the “BUPT‑2” and “BUPT‑3” satellites for nearly a year, boosting image‑processing efficiency by a factor of 50 while also validating the reliability of a 6G onboard core network. The world’s first space‑based computing cloud is now in regular operation, serving over a hundred users with an inference energy efficiency of 10 tokens per joule. Shanghai Dongfang TianSuan Technology and Guangbenwei Intelligent Technology have jointly initiated the development of the world’s first space‑based optical computing satellite. At the national level, in April 2026, the China National Space Administration will lead the launch of a special feasibility study for a space‑based intelligent computing constellation.
The core value of “TianShu TianSuan” lies in compressing the decision‑making pipeline down to the minute level, addressing the absurd reality under the traditional “sky‑sensing, ground‑computing” paradigm, where 95% of space‑borne data is discarded due to bandwidth constraints. However, engineering challenges remain: in a vacuum, heat dissipation relies on radiation, necessitating breakthroughs in liquid cooling and passive radiative cooling technologies; limited solar‑array power imposes constraints on scaling high‑power computing capabilities; inter‑satellite high‑speed communication depends on the maturity of laser‑based networks; and commercialization must focus on high‑value‑added applications—such as emergency relief, offshore monitoring, and national defense—that are highly sensitive to response speed. This launch marks the starting point of the “XingSuan” constellation‑building plan, which aims to deploy 2,800 satellites by 2030 for commercial use and achieve full network coverage by 2035, propelling satellites from passive sensors to autonomous, decision‑making space‑borne intelligent agents.