CGWS(SZ:002939) Xin Lang Cai Jing·2025-12-18 04:07Core Insights - Traditional data centers face challenges in high-parallel intelligent computing scenarios, leading to the emergence of AIDC with a new architecture featuring a computing power base layer, platform service layer, model enabling layer, and industry application layer [1][4] - AI servers are projected to consume five times more power than traditional servers, resulting in a doubling of total power consumption in data centers by 2030, with the US, China, and Europe accounting for 82% of global data center capacity [1][4] Group 1: AIDC Architecture - AIDC requires a computing power base to be constructed as a highly coordinated system akin to a supercomputer, moving towards "DC as a Computer" [1][4] - The architecture allows for decoupling between layers while requiring high vertical integration within each layer to ensure efficient computing power and integrated operations [1][4] Group 2: Power Supply Solutions - The transition to HVDC technology is necessary to support future data center power demands, with NVIDIA leading the shift to 800V HVDC infrastructure starting in 2027 [2][5] - The global market for AI data center 800V HVDC is expected to reach 35.4 billion yuan by 2030, with a compound annual growth rate of 46% from 2026 to 2030 [2][5] Group 3: Cooling Solutions - As AI technology evolves, cooling solutions are diversifying, with liquid cooling becoming essential as rack power density exceeds 40-60 kW [3][6] - AI data center IT energy consumption is projected to reach 77.7 TWh by 2025, doubling from 2023, and further increasing to 146.2 TWh by 2027, with a compound annual growth rate of 44.8% from 2022 to 2027 [3][6] - The liquid cooling market in China is expected to reach 10.5 billion USD by 2028, with a compound annual growth rate of approximately 48.3% from 2023 to 2028 [3][6]