Sou Hu Cai Jing·2026-03-08 18:23Core Insights - The report from Huatai Securities quantifies the impact of AI Token deployment on China's power industry, indicating that the transition to the reasoning era in AI could lead to a 10% elasticity in electricity demand, boosting green certificates and capacity prices [1][2]. Group 1: AI Industry Transition - The AI industry has shifted from a training era to a reasoning era, with a narrowing gap in computing power between domestic and overseas players. The Agent model is expected to drive exponential growth in Token consumption [1][2][9]. - If the global daily Token call volume reaches trillions, combined with a 30%-50% market share of domestic large models and 70%-90% local computing power deployment, Token deployment could increase China's electricity and power demand by 8% and 18%, respectively [1][2]. Group 2: Electricity Cost Dynamics - The importance of electricity costs in AI computing competition is increasing, with the share of electricity in unit Token costs rising significantly. In high-performance training versions of AIDC, electricity accounts for only 5%, but this doubles to 10% under reasoning models, and can reach 20%-30% with self-developed reasoning-grade chips [1][7][9]. - The report highlights that while the current electricity cost is only 10% of Token costs, this share is expected to continue rising as chip efficiency improves [9][18]. Group 3: Price Elasticity and Market Dynamics - The demand for Tokens is expected to enhance China's green electricity demand by 4%-33% from 2026 to 2030, benefiting undervalued green certificate prices. The low utilization rate of reasoning models is likely to increase capacity prices by 50-300 yuan per kilowatt during the same period, while the impact on electricity prices will be relatively delayed [2][8]. - The report contrasts with market views by emphasizing that the AI race has entered the reasoning era, and the elasticity of Token demand on green certificates and capacity prices is significantly higher than on electricity prices [2][9]. Group 4: Investment Recommendations - The report recommends focusing on undervalued stocks in the green and thermal power sectors, particularly those benefiting from renewable energy demand, such as Longyuan H, Green Development, and China Power [10]. - Companies like Jinko Power, Jingneng Clean Energy, and others are highlighted for their potential to benefit from capacity price elasticity [10]. Group 5: Future Outlook - The report suggests that the power supply in China will not become a bottleneck for computing power expansion, given the country's ample electricity supply. The industrial electricity price gap between China and the U.S. is expected to further highlight China's advantages in power supply [1][7][21]. - The transition to the reasoning era is anticipated to attract more infrastructure investments, as the sensitivity of electricity costs in AIDC is expected to double, making it a more critical factor in the competitive landscape [20][21].