Workflow
国产算力价值量拆分
2025-08-18 01:00

Summary of Key Points from the Conference Call Industry Overview - The domestic computing power industry chain includes AI chips (e.g., Cambricon, Haiguang), AI servers (e.g., Inspur), server power supplies (e.g., Oulu Tong), and computing power leasing (e.g., Fang Technology, Hongjing Technology) [1] - The market space for the AI industry chain is closely related to the power consumption and quantity of AI chips [1] Core Insights and Arguments - Capital expenditure primarily comes from AI giants (ByteDance, Alibaba, Tencent) and AIDC companies (Guanghuan New Network, Aofei Data) [1][4] - Example calculation: 1 million AI chips with an average power consumption of 700 watts correspond to 120,000 servers, with a total power consumption of 1.4GW [1] - Total value of AI chips is approximately 110 billion yuan, AI servers around 150 billion yuan, and server power supplies about 3.2 billion yuan [1][5] - AIDC companies' total capital expenditure is around 28 billion yuan, with 40%-50% allocated to power distribution systems and 15%-20% to cooling systems [1][6] Liquid Cooling Industry - The liquid cooling industry is accelerating and has significant potential, with market size estimates reaching hundreds of billions to over a trillion yuan [1][8] - Liquid cooling market growth is expected to outpace domestic growth, with the overseas market projected to reach at least hundreds of billions by 2026 [8] - The penetration rate of liquid cooling is currently low, but it is expected to expand as the number of domestic chips increases [8] Network Equipment Demand - Demand for network equipment must consider the network architecture of major companies, with specific ratios for GPU optical modules [9] - Accurate calculations of optical module and switch demand require understanding each company's network structure [9] Global AI Industry Trends - Optimistic outlook for the global AI industry chain, with expectations for valuation increases and a turning point for the domestic AI industry chain [10] - Key sectors to watch include optical communication (e.g., Zhongji Xuchuang, Yuanjie Technology) and liquid cooling [10] Companies to Watch - Key companies in the domestic computing power sector include Cambricon, Haiguang, Oulu Tong, and Inspur [11][12] - Monitoring the progress of NVIDIA H20 and RTX Pro cards is crucial, as successful sales could benefit the entire domestic AI industry chain [12] - Other notable companies include Guanghuan, Aofei, Zhongheng Electric, Kehua Data, and Yingwei Technology, which are positioned to benefit from industry growth [12]