China Tightens Data-Center Build Controls as Regions Compete on Power Incentives
DataNews Editorial Team
Chinese regions are using discounted electricity, subsidies, tax and lease concessions, and compute vouchers to attract data-center and AI workloads under the East Data, West Computing strategy. At the same time, authorities are restricting new construction in regions where existing data-center utilisation is below 50%.
- Chinese regions are competing for AI and data-center investment through electricity discounts and other incentives.
- Western China is being positioned for energy-intensive workloads under the East Data, West Computing strategy.
- New construction is being restricted where existing data-center utilisation is below 50%.
- Reported tariff advantages do not remove the importance of grid connectivity, latency, demand proximity and realised utilisation.
- Beijing’s reported tariff reduction of up to 50% applies only to an unspecified subset of large facilities using domestic AI chips.
China’s regional competition for data-center and AI infrastructure is increasingly centred on energy cost. Under the East Data, West Computing strategy, energy-intensive data-center workloads are being encouraged to move to western regions with cooler climates and more accessible renewable-energy resources.
According to the available source, provinces and municipalities are using lower electricity tariffs alongside subsidies, tax and lease concessions, and compute-capacity vouchers to secure projects. In selected distributed solar projects in western China, electricity prices were cited at CNY0.19/kWh, compared with CNY0.43/kWh in eastern regions.
Beijing has reportedly cut electricity tariffs by as much as 50% for some large data centers using domestic AI chips. The source does not identify the facilities involved, the eligibility criteria or the duration of the concession.
The central policy approach is also becoming more selective. New data-center construction is being limited in regions where utilisation of existing facilities is below 50%, reflecting concern over duplicated development and underused capacity. The source separately cites utilisation of only 20–30% at some unnamed facilities in Europe and the United States, against a stated policy target of 60%; it provides no site-level detail, measurement period or methodology.
Shaanxi, which is not among the official East Data, West Computing clusters, has built 22 large data centers and three data-industry parks, according to the source. No capacity, commissioning dates or current operating utilisation were provided.
For operators and investors, the policy direction suggests that subsidised power could accelerate the geographic redistribution of AI workloads toward energy-surplus regions and support demand for local compute infrastructure. However, the construction restrictions indicate a parallel risk of excess capacity where siting decisions are driven primarily by tariffs rather than grid connectivity, proximity to demand and actual workload utilisation.
The regional contrast is also evident in Vietnam, where operators Viettel and CMC reported that electricity costs rose by more than 50% during the first three billing periods after a new tariff was introduced. The example underscores the sensitivity of data-center operating economics to power-price policy.
The policy mix makes electricity pricing a direct tool of regional competition for AI and data-center infrastructure. It also signals that low power costs alone may not support viable operations: transmission infrastructure, demand proximity, latency and realised utilisation remain critical to project economics.
- — / September 23, 2026 · JapaneseSource →
This article was prepared by the DataNews editorial team based on the sources listed above.