Перейти к содержимому
Новости

Ulanqab Spotlighted as China Links Data-Center Growth to Power and Water Constraints

Редакция ЦОД:Главное

Источник изображения: первоисточник

During an inspection trip to Ulanqab, Inner Mongolia, Xi Jinping stressed the need to coordinate computing infrastructure with electricity and water resources in the construction and operation of computing centers. The emphasis places power availability, cooling-water constraints and operating efficiency at the center of expansion planning for the city’s AI-compute cluster.

Главное
  • China’s leadership is explicitly linking compute-cluster development with both electricity and water-resource planning.
  • Ulanqab’s reported 21 GW of new-energy capacity is a citywide generation metric, not a verified supply allocation to data centers.
  • Water scarcity may constrain water-intensive cooling expansion despite Ulanqab’s cold-climate and renewable-energy advantages.
  • Reported computing-scale figures use an undefined unit, “P,” and cannot be converted reliably into MW or FLOPS from the available information.
  • The reported CNY500 billion-plus investment figure does not separate deployed CAPEX from announced, construction-stage or planned investment.

Xi Jinping’s inspection visit to Ulanqab brought renewed attention to “computing-power coordination” and “computing-water coordination” in data-center development. The stated direction is for computing centers to be built and operated efficiently, sustainably and safely, with computing infrastructure planned alongside electricity and water supply.

Ulanqab is described as a rapidly expanding AI-compute cluster. According to the cited publication, aggregate investment in the city’s computing capacity exceeds CNY500 billion, although the figure does not distinguish between capital already deployed and investment associated with announced, under-construction or planned projects.

The publication reported computing scale of 125,000 P at the end of 2025, including more than 61,000 P of “green computing” supplied to Beijing. By mid-October 2026, it said Ulanqab’s computing scale had reached 190,000 P, with intelligent computing accounting for more than 95%. The unit “P” is not defined in the source and should not be treated as directly comparable with MW, FLOPS or another standardised capacity metric.

Power-resource availability is central to the cluster’s positioning. Ulanqab reportedly has 21 GW of installed new-energy capacity, equivalent to 67% of the city’s total installed generating capacity. This is a citywide power-system measure, however, rather than evidence of the actual energy mix serving individual data centers.

Water remains the principal qualification to the power-and-climate case. The publication notes that Ulanqab lacks major rivers and identifies water availability as a consideration for further data-center cooling expansion. This could increase the value of sites that combine clean generation, favourable ambient conditions and demonstrable water security, while making cooling technology selection, water-use permissions and environmental constraints more material to AI-data-center competitiveness.

The source also attributes two policy-related benchmarks to new data centers: an 80% green-electricity share and PUE below 1.2. It does not identify the underlying regulatory document, applicability or effective date, so these figures should be treated as reported benchmarks rather than confirmed requirements. Separately, the publication states that the first phase of Huawei’s Ulanqab data center was completed in 2016.

Почему это важно

The event underscores that low-carbon power and water availability are becoming fundamental siting and operating constraints for large AI data-center clusters in China. Ulanqab may demonstrate the advantages of locating workloads near cool climates and substantial renewable-energy resources, but water scarcity could limit the expansion of water-intensive cooling designs.

Источники

Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.