电力AI行业跟踪（一）：探讨DeepSeek对AI电力需求的影响

Investment Rating - The report maintains a "Buy" rating for the power equipment sector, indicating an expected investment return exceeding 15% over the next 6-12 months [5][8]. Core Insights - The emergence of high-efficiency AI models like DeepSeek raises concerns about short-term growth in AI power demand, as these models significantly reduce the computational resources required while maintaining or enhancing performance [1][2]. - Despite short-term fluctuations, the long-term trend for AI power demand is expected to remain upward, driven by factors such as increased GPU demand due to rapid AI development, accelerated deployment of AI applications, and sustained capital expenditure growth from tech giants [3][4]. Summary by Sections AI Power Demand Trends - AI power demand has seen rapid growth, with U.S. data center electricity consumption rising from 76 TWh in 2018 to 176 TWh in 2023, projected to reach 325-580 TWh by 2028, accounting for 6.7%-12.0% of total U.S. electricity consumption [3]. - The introduction of efficient models like DeepSeek may temporarily impact computational demand but is expected to lower overall computational costs, encouraging further investment in AI model development [2][3]. Investment Opportunities - The report suggests focusing on leading domestic server power manufacturers with strong R&D capabilities, such as Megmeet and Euron, as well as companies benefiting from NVIDIA server upgrades like Blue Lithium and Jianghai [4]. - Companies with established HVDC products and development capabilities are also highlighted as potential beneficiaries, including Hezhong Electric, Zhongheng Electric, and Kehua Data [4].