AIDC：人工智能发展的核心引擎产业赛道投资图谱

Group 1: AIDC Overview - AIDC is a computing infrastructure built on artificial intelligence computing architecture, integrating high-performance computing equipment, high-speed networks, and advanced software systems to provide an efficient and stable computing environment for AI training and inference [1][10] - The core components of AIDC include computing devices, storage devices, network devices, power supply systems, and cooling systems, which collectively support powerful computing capabilities, efficient data processing, and various AI applications [10][11] Group 2: Market Trends and Policy Support - The AIDC market is expected to grow significantly, with an annual growth rate exceeding 25% from 2023 to 2028, leading to a market size of over 280 billion yuan by 2028 [16] - AIDC's policy support features three main characteristics: multi-departmental collaboration, full-cycle coverage, and a focus on technological self-sufficiency, with various policies established since 2017 to strengthen AI data center development [14][15] Group 3: AIDC Market Performance - AIDC has experienced three significant phases of excess returns from 2018 to 2025, driven by policy initiatives, technological breakthroughs, and the competitive landscape of AI models [2][27] - The first phase (2019-2020) saw a cumulative increase of 110% in computing infrastructure, while the second phase (2022-2023) recorded a 50% increase, and the upcoming phase (2024-2025) is expected to be driven by heightened demand for computing power due to AI model competition [2][24][25] Group 4: Industry Chain Overview - In the AI computing chip sector, GPUs dominate the market with a share of approximately 90%, and the market size for GPUs in China is projected to exceed 120 billion yuan by 2025 [30][34] - The storage chip market, supported by DRAM and NAND flash, is entering a recovery phase driven by the increasing demand for high-bandwidth storage due to AI model training [42][43] - The power supply and cooling systems are also evolving to meet the rising demands of AI computing, with opportunities in modular and prefabricated solutions for power supply and liquid cooling technologies becoming more prevalent [4][5][29] Group 5: Business Models in AIDC - AIDC encompasses four core business models: IaaS (Infrastructure as a Service), PaaS (Platform as a Service), SaaS (Software as a Service), and MaaS (Model as a Service), catering to various customer needs from infrastructure hosting to customized AI model services [22][26] - The diverse stakeholders in the AIDC ecosystem include local governments, internet and cloud service providers, third-party IDC service providers, and AI enterprises, each with distinct strategic goals and resource advantages [19][20]