Chinese President Xi Jinping wants computing power to behave more like electricity. Under that model, it would be generated where it is cheap, moved across the country, and consumed wherever demand happens to be.
Speaking in Inner Mongolia, Xi called for building a national integrated computing power network and for deeper integration of the digital economy with the real economy.
Why Inner Mongolia keeps winning the data centers
Electricity costs in the region average around 0.33–0.35 yuan per kWh, backed by abundant renewable energy. Connections from Inner Mongolia to major economic centers like Beijing run at under 5 ms of latency.
As of May 2026, Inner Mongolia’s computing power totaled 315,000 petaflops, roughly one-seventh of China’s national capacity. Of that, 297,000 petaflops counts as intelligent computing, the category built for AI training and inference.
As of the first half of 2026, 20 of China’s 30 leading private computing companies were operating in Inner Mongolia, with investment exceeding 200 billion yuan.
Ulanqab has signed agreements for 89 data center projects. Of those, 86 are AI-focused, with total computing capacity of approximately 15 GW.
State-owned China Mobile has also expanded its footprint across Hohhot and Ulanqab. Its computing center capacity now reaches 20,100 petaflops, with 16,700 petaflops dedicated to AI and smart computing applications.
The company’s expanded Hohhot center, completed in October 2026, directs 90% of its capacity to the Beijing-Tianjin-Hebei region. It runs on 88% green electricity.
East Data, West Computing, five-year-plan edition
Xi’s call builds on the ‘East Data, West Computing’ strategy, formally approved in 2021, which shifts heavy processing to western and northern regions with spare land and power while keeping data generation in China’s wealthy eastern cities.
The push also sits inside the ‘six networks’ framework laid out in China’s 15th Five-Year Plan, covering 2026 to 2030, which aims to build a coordinated system of computing hubs and clusters.
China aims to scale intelligent computing to 9,800 EFLOPS by 2030. The plan is supported by 70 transmission corridors and 17 approved interconnection nodes.
What this means for China’s AI buildout
China Mobile is the most concrete example of capital deployment, given its expanded Hohhot and Ulanqab facilities and its role routing 90% of capacity to the Beijing region. Other large players tied to AI and cloud services in China, including Huawei, ByteDance and Alibaba, have been linked to the broader expansion.
Ulanqab’s roughly 15 GW of committed capacity still has to be constructed, powered, and filled with customers. The 17 approved interconnection nodes are the plumbing that will determine whether this becomes one network or many isolated islands.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.