After three years of rapid expansion of its digital infrastructure, China is taking steps to clean up its rapidly growing data centre market. The central government is currently conducting a nationwide assessment of the sector, which aims to curb excess computing capacity, counteract declining efficiency and create a more coherent system for managing these resources.
Behind the data centre boom was, among other things, the ‘Eastern Data, Western Computing’ programme, which promoted the construction of facilities in the western regions of the country. Cheap land and low energy costs were supposed to translate into efficiency, but in practice investments were often disconnected from real market needs. The result is hundreds of inefficient centres, many of which use only 20-30% of their capacity. More than 100 projects have been cancelled in the last 18 months, signalling growing uncertainty about the viability of these projects.
In response, Beijing is working to develop a nationwide cloud platform to combine distributed computing resources into a single, more flexible structure. The project, coordinated by the Ministry of Industry and Information Technology in collaboration with three state-owned telecommunications operators, aims to harness underutilised resources and make them more accessible at the national level.
In parallel, the National Development and Reform Commission is introducing new regulations to prevent inefficient investments. Local authorities have been banned from participating in smaller infrastructure projects, and new initiatives must meet certain efficiency thresholds and have secured contracts for the purchase of computing power.
Implementing such a network, however, is a huge technological challenge. Connecting data centres into a coherent structure requires not only stable low-latency connections, but also the integration of different hardware and software architectures, which is made difficult by, among other things, local alternatives to Nvidia chips.
Despite the difficulties, Beijing sees the project as a key step towards greater technological sovereignty and more sustainable development of the AI sector. However, the first effects of these measures will not be seen until after 2028.
