A recent research report indicates that as AI chip and server power density continues to climb, Vertiv (VRT.US) projects that by 2029, individual AI GPU cabinet power will exceed 1MW, with single AI POD power surpassing 500kW, presenting significant challenges for cooling systems. When rack density rises to 20kW, liquid cooling advantages become prominent. According to IDC, China's liquid cooling server market reached $2.37 billion in 2024, up 67.0% year-over-year, with projected CAGR of 46.8% from 2024-2029. Global AI server liquid cooling market is estimated to reach $3.07 billion in 2025 and $8.58 billion in 2026. Focus is recommended on leading system solution providers, core component manufacturers, and server vendors actively deploying liquid cooling technology.

**Key Investment Themes:**

**Rising Chip and Server Power Density Makes Liquid Cooling Replacement of Air Cooling an AI Era Trend**

AI chip and server power density is escalating rapidly. Vertiv forecasts that by 2029, single AI GPU cabinet power will exceed 1MW, with single AI POD power surpassing 500kW, creating substantial cooling system challenges. When rack density increases to 20kW, liquid cooling advantages become evident. Additionally, as data centers drive up electricity demand and PUE becomes a regulatory focus, cooling systems represent the second-largest energy consumption center in data centers, making cooling efficiency improvements critical. Liquid cooling offers superior heat dissipation efficiency, enhanced computing power deployment density, and reduced system power consumption, positioning it to replace traditional air cooling.

**Cold Plate Liquid Cooling Leads Industrialization, While Immersion and Spray Cooling Remain in Early Technical Exploration**

Based on whether server internal components directly contact cooling fluid, liquid cooling can be categorized into indirect contact (cold plate liquid cooling) and direct contact (immersion and spray liquid cooling). Cold plate liquid cooling systems allow cold plates to directly attach to chips and other server components for heat dissipation. Compared to direct contact liquid cooling, this approach requires minimal modifications to equipment and facility infrastructure while providing significant cooling improvements over air cooling, making it the most mature liquid cooling solution currently available. Cold plate liquid cooling is expected to achieve large-scale deployment first.

**Industry Chain Competition Intensifies with Broad Market Opportunities**

The industry chain is experiencing accelerated deployment across upstream (cold plates, CDUs, quick connectors), midstream (server manufacturers, solution providers), with downstream cloud providers leading large-scale applications. NVIDIA's GB200 and GB300 NVL72 are accelerating ecosystem formation, creating vast market opportunities. According to IDC, China's liquid cooling server market reached $2.37 billion in 2024, growing 67.0% year-over-year, with projected CAGR of 46.8% from 2024-2029. Analysis estimates global AI server liquid cooling market could reach $3.07 billion in 2025 and $8.58 billion in 2026.

**Investment Targets**

Focus is recommended on leading system solution providers, core component manufacturers, and server vendors actively deploying liquid cooling technology. Related industry chain targets include Vertiv (VRT.US), Siquan New Materials (301489.SZ), Golaub (300499.SZ), Shenling Environment (301018.SZ), Airong (300602.SZ), Sugon Digital (872808.BJ), Sobede (300322.SZ), Jiebang Technology (301326.SZ), Qihong Technology, Dual Hong, Delta, BYD ELECTRONIC (00285), FIT HON TENG (06088), InVIDEC (002837.SZ), Inspur Information (000977.SZ), Tsinghua Unigroup (000938.SZ), LENOVO GROUP (00992), Dawning Information (601138.SH), Industrial Fulian (601138.SH), ZTE Corporation (000063.SZ), among others.

**Risk Factors**

AI application deployment progress and liquid cooling penetration rates may fall short of expectations, while new technology developments could reshape the industry chain.