Ge Long Hui·2025-11-04 09:45Core Insights - The critical bottleneck for AI development is electricity supply, as highlighted by Microsoft CEO Satya Nadella, emphasizing that without power, high-performance GPUs remain idle [1][3] - The total capital expenditure for AI in China is projected to reach 2-2.5 trillion yuan by 2030, with 800 billion yuan allocated for non-IT infrastructure like power and cooling [1][3] Short-term Opportunities (1-3 years) - The demand for AI Data Center (AIDC) power and electrical equipment is expected to surge, as these are essential for AI operations [4] - AI servers now require significantly more power, with configurations moving from two 800W power supplies to four 1800W supplies, indicating a doubling of power needs [7] - High-efficiency power supplies are becoming a necessity, with companies like 欧陆通 and 麦格米特 leading the market with high-capacity offerings [8][9] Mid-term Opportunities (3-5 years) - Solid State Transformers (SST) are anticipated to become a key component in AIDC power supply, with a projected market size reaching hundreds of billions by 2030 [11][14] - The efficiency of SSTs is significantly higher than traditional transformers, making them ideal for high-power data centers [13] - Major companies are already investing in SST technology, indicating strong future demand [19] Long-term Opportunities (5-10 years) - Thorium molten salt reactors and nuclear fusion are seen as revolutionary solutions for stable and clean energy, crucial for supporting the growing power demands of AI [23][26] - Thorium reactors are expected to be commercially deployed by 2030, with a total market size projected to reach trillions [23] - The rapid advancements in nuclear fusion technology could potentially eliminate power constraints for AI, with commercial viability expected post-2040 [26]