The Zhitong Finance App learned that the Shanghai Stock Exchange released a research report saying that large models of artificial intelligence and smart applications are driving continuous growth in computing power demand. In 2025, China's intelligent computing power scale will reach 1,590 EFLOPS, and the target scale for 2030 is 9,800 EFLOPS. Liquid cooling technology can reduce refrigeration system energy consumption and improve data center PUE through high liquid heat transfer efficiency. The bank believes that under the wave of intelligent computing power construction, green, low-carbon and sustainable development have become important guidelines for computing power infrastructure. The value of liquid cooling technology continues to be highlighted. It is recommended to focus on the two major directions of liquid cooling systems and core components, liquid cooling servers, and complete cabinets.
The main views of Shanghai Securities are as follows:
Policies require two-wheel drive, intelligent computing to move towards green and low-carbon
Large models of artificial intelligence and smart applications drive continuous growth in computing power demand. In 2025, China's intelligent computing power scale will reach 1,590 EFLOPS, and the target scale for 2030 is 9800 EFLOPS. Demand for electricity in computing power centers is rising at the same time. In 2025, the electricity consumption of China's computing power centers will reach 19.6 billion kilowatt-hours. At the same time, national and local authorities continue to increase energy efficiency requirements for data centers. The “Special Action Plan for Green and Low-Carbon Data Center Development” proposes that by the end of 2025, the PUE for the construction, renovation and expansion of large and very large data centers will not be higher than 1.25, and the PUE for national hub node data center projects will not be higher than 1.2.
Liquid cooling technology is deeply empowered to reshape the computing power cooling system
The power consumption of AI chips continues to increase, and the increase in the power density of a single cabinet places higher demands on the cooling capacity of data centers. Liquid cooling technology can reduce refrigeration system energy consumption and improve data center PUE through high liquid heat transfer efficiency. Among them, cold plate liquid cooling is highly compatible with existing air cooling systems, and the difficulty and cost of transformation are relatively manageable; immersion liquid cooling has heat dissipation advantages in high heat flow density scenarios. TrendForce Jibang Consulting predicts that the penetration rate of liquid cooling in AI chips will reach 53% in 2026; CCI Consulting predicts that China's liquid cooling data center market will grow from 23.25 billion yuan in 2026 to 47.04 billion yuan in 2028.
Benchmarking projects are implemented at an accelerated pace, and the practical value of the industry is highlighted
Liquid cooling technology has been applied in a number of large-scale intelligent computing projects. China Mobile's Hohhot Intelligent Computing Center deployed 2,390 liquid-cooled intelligent computing servers, with the PUE as low as 1.15; China Telecom's Shaoguan Wanka Cluster deployed 30 domestic supernodes, fully using cold plate liquid cooling, and completed construction, commissioning and launch in just 90 days; Inspur Information Element Computing Power Plant raised the maximum load limit of liquid cooled cabinets to 130 kW, and the PUE can be reduced to less than 1.1. The bank believes that under the wave of intelligent computing power construction, green, low-carbon and sustainable development have become important guidelines for computing power infrastructure. The value of liquid cooling technology continues to be highlighted. It is recommended to focus on the two major directions of liquid cooling systems and core components, liquid cooling servers, and complete cabinets.
Risk warning: Downstream prosperity falls short of expectations; policy advancement falls short of expectations; technological innovation falls short of expectations, and industry competition intensifies.