电力设备与新能源行业：“AIDC系列报告”：关注电源+供配电系统新趋势

Investment Rating - The report maintains a "Buy" rating for the electric power equipment and new energy industry [4]. Core Viewpoints - The AI-driven demand for computing power is expected to significantly increase electricity consumption, with AIDC (Artificial Intelligence Data Center) becoming a core component of this trend [10][11]. - The transition from traditional data centers to AIDC is driven by the need for enhanced resource management and flexibility to support AI applications [16]. - The report highlights the importance of power supply and distribution systems in AIDC, emphasizing their role in ensuring stable and reliable power for high-density computing environments [25][26]. Summary by Sections 1. Computing Power and Electricity Demand - AI's rapid development necessitates substantial computing resources, which in turn rely heavily on electricity supply [10]. - The energy consumption of AI models, such as ChatGPT, is significantly higher than traditional computing tasks, indicating a growing demand for electricity in data centers [10]. 2. AIDC and Power Supply Systems - AIDC centers are designed to meet the specific needs of AI computing tasks, differing significantly from traditional data centers in architecture and cooling methods [16][17]. - The global data center IT power demand is projected to surge from 49 GW in 2023 to 96 GW by 2026, with AIDC contributing approximately 40 GW of this increase [17][18]. 3. AI Server Power Supply - The report discusses the evolution of power supply systems in data centers, highlighting the shift from UPS to HVDC systems for improved efficiency [36]. - AIDC's power density is expected to rise dramatically, with predictions of cabinet power levels reaching up to 100 kW by 2029 [44][45]. - The report identifies key players in the power supply market, noting that Taiwanese companies dominate, with Megmeet entering the NVIDIA supply chain [54][55].