China Expands Liquid Cooling Deployment to Curb Soaring Data Center Power Use

As artificial intelligence workloads drive server heat and electricity demands to new highs, Chinese data centers are rapidly shifting toward liquid cooling technologies. In national computing hubs such as Guizhou and Ningxia, operators are deploying immersion and cold-plate liquid cooling modules paired directly with local renewable power. Driven by national policy guidelines targeting computing energy efficiency and manufacturer commercialization, these systems reduce power usage effectiveness (PUE) significantly below traditional air cooling, though upfront costs and operational integration challenges persist.

Cooling is no longer just auxiliary infrastructure; it is becoming a primary bottleneck for scaling high-density AI clusters. Liquid cooling drastically cuts the overhead power required for thermal management, allowing operators to meet stringent carbon and efficiency standards while expanding compute capacity.

As the parameter sizes of large artificial intelligence models expand and high-performance computing hardware accelerates, server thermal output is posing an acute operational challenge for data center operators across China. According to a report by the People's Daily, thermal dissipation is driving a major infrastructure transition from conventional air conditioning to advanced liquid cooling solutions. In the Guian New Area of Guizhou province, the Southern Energy Big Data Center has deployed a 40-foot containerized immersion liquid cooling compute pod. According to the project lead, conventional air cooling requires an additional 0.3 to 0.5 kilowatt-hours of electricity for thermal management for every kilowatt-hour spent on computing tasks. By submerging servers into specialized dielectric coolant, this additional overhead is reduced to approximately 0.15 kilowatt-hours. The pod connects to an integrated power-carbon-compute dispatch system that dynamically adjusts computing loads based on local hydropower output. Policy mandates are accelerating this deployment. The China Academy of Information and Communications Technology noted in its 2026 green computing research report that liquid cooling has become an indispensable requirement for high-density computing centers. Furthermore, four government agencies, including the National Development and Reform Commission, issued an action plan to promote bidirectional enablement between AI and energy, emphasizing continuous improvements in the energy and carbon efficiency of computing facilities. An accompanying implementation plan for energy-saving equipment spanning 2026 to 2028 targets the development of high-performance thermal materials and phase-change liquid cooling. Hardware vendors are reporting increased order deliveries. According to public disclosures and mid-2026 financial reports, cooling equipment makers including Envic, Gaolan, and Fnode are scaling volume deliveries of cold-plate and immersion modules. Regional projects illustrate diverse deployment models: Ningxia's Zhongwei hub combines full natural air cooling with localized liquid cooling to run on 100 percent renewable power with an annual power usage effectiveness below 1.2, while China Mobile has deployed large-scale cold-plate systems in the Yangtze River Delta, and Sugon has demonstrated phase-change immersion setups reaching a PUE as low as 1.03. Industry analysts note, however, that widespread adoption faces bottlenecks. High upfront capital costs, incomplete operations and maintenance standards, and unproven business models for waste heat recovery remain obstacles, while some projects still struggle to fully coordinate cooling hardware with real-time green power grid dispatching.