Changjiang Securities·2026-02-24 11:28Investment Rating - The report maintains a "Positive" investment rating for the electrical equipment industry [12]. Core Insights - The report highlights that the increasing demand for AI is exacerbating the electricity shortage in the U.S., leading to rising electricity prices and a shift towards self-built power sources for data centers [4][20]. - Legislative actions are accelerating the implementation of policies that encourage data centers to build their own power sources, which is expected to significantly boost the demand for transformers and energy storage systems [7][33]. - The report quantifies the expected growth in transformer demand due to self-built power sources, estimating a compound annual growth rate (CAGR) of approximately 38.4% for main transformers and 15.4% for distribution transformers from 2026 to 2030 [54][57]. Summary by Sections Self-Built Power Sources for Data Centers - The report indicates that self-built power sources for data centers in the U.S. are becoming a prevailing trend due to the increasing electricity demand and regulatory pressures [7][20]. - It discusses the anticipated power supply architecture for data centers, which will require higher voltage levels (230-500kV) to accommodate larger capacities [8][34]. Transformer Demand - The report estimates the transformer demand for self-built power sources, projecting that from 2026 to 2030, the demand for main transformers will grow significantly, with specific capacity requirements outlined for different voltage levels [9][52]. - The estimated demand for transformers includes 10.59 GW for 230kV, 14.82 GW for 138kV, and 26.32 GW for 34.5kV in 2026, with substantial increases expected in subsequent years [52]. Energy Storage Demand - The report analyzes the energy storage requirements for data centers, projecting a compound annual growth rate of 23% for energy storage needs, potentially exceeding 150 GWh by 2030 [10][58]. - It outlines various applications for energy storage, including supporting power supply, demand response, and load smoothing, which are critical for managing the fluctuating demands of AI workloads [60][63].