财经观察：“人工智能雄心”冲击美国电网

Core Insights - The rapid expansion of artificial intelligence (AI) by large tech companies is significantly impacting the U.S. power grid, revealing its vulnerabilities and leading to increased electricity costs for consumers [1][5]. Data Center Growth - As of Q2 2023, the U.S. has approximately 522 large-scale data centers, accounting for about 55% of global computing power, with an expected addition of around 280 more by the end of 2028 [2]. - The energy consumption of a single data center can be equivalent to that of 1,000 Walmart stores, and an AI search consumes ten times more electricity than a Google search [2]. Electricity Demand Projections - Data centers' electricity consumption was less than 2% of total U.S. electricity usage before 2020, but it is projected to rise to 12% by 2028 [3]. - Deloitte forecasts that the electricity demand from AI data centers could increase by over 30 times by 2035 compared to 2024 levels [3]. Infrastructure Challenges - The U.S. requires an additional 50 gigawatts of electricity capacity to support the rapid growth of AI, which is enough to power approximately 40 million households [4]. - Upgrading the aging power infrastructure is expected to cost utilities around $1.1 trillion from 2025 to 2029, with costs likely to rise due to tariffs on materials [4]. Regional Power Strain - Areas with concentrated AI industries, such as Texas and Georgia, are experiencing significant pressure on their power grids, leading to increased electricity prices and potential shortages [5][8]. Rising Electricity Prices - The average electricity price in the U.S. has increased by 13% since 2022, primarily due to rising costs associated with updating and maintaining the power grid [6][8]. - In Q3 2023, the average electricity and gas bills for consumers rose by 3.6% year-over-year, attributed to the surge in demand from AI developments [7][8]. Self-Supply Initiatives - Some tech companies are opting to build their own power plants to meet the growing demand, but the existing grid's transmission and access capabilities remain a significant bottleneck [9][10]. Policy and Future Outlook - The current U.S. administration's focus on fossil fuels may provide short-term relief for electricity supply pressures, but the pace of infrastructure development may not keep up with the rapid growth in AI demand [11].