China's Eight National AI Computing Hubs Move Toward Commercial Operation
China’s eight national computing hubs, launched in 2022, now include ten planned data-center clusters across regions such as Beijing-Tianjin-Hebei, the Yangtze River Delta, the Greater Bay Area, Inner Mongolia, Guizhou, Gansu, and Ningxia. CCTV reported that Zhangjiakou has kept one-way network latency below 2 milliseconds, Horinger’s green-power share exceeds 85%, and Qingyang is developing into a base for large-model training and token computing. TMTPost also highlights the growing need to prove commercial viability through utilization, customers, revenue, and investment returns.
The update shows how China’s national computing infrastructure is adapting from large-scale construction toward AI training, inference, and cross-region service, while making utilization and commercial demand central to the hubs’ competitiveness.
(本文作者为 文武赵,钛媒体经授权发布)
文 | 文武赵
2026年10月28日,央视新闻报道了中国八大国家算力枢纽的最新建设进展。
北京周边的张家口,单向网络时延已经控制在2毫秒以内;内蒙古和林格尔,绿电占比超过85%;甘肃庆阳,正成为国内大模型训练和词元计算的重要基地。
从张家口到芜湖,从贵州贵安到宁夏中卫,一张横跨中国东西部的算力网络逐渐形成。
一个值得讨论的问题是,为什么偏偏是这些地方?
#National Development and Reform Commission#National Data Administration#Zhangjiakou#Horinger#Qingyang#Wuhu#Shaoguan#Gui'an#Zhongwei#Tianfu#Chongqing#Zhangjiakou–Wuhu–Gui'an–Zhongwei computing network