2024-09-09-EPRI-为智能赋能—人工智能和数据中心能耗分析（英）_35页_3mb

报告摘要

Data Center Growth: Global data centers are among the fastest-growing industries, with significant expansion in the United States and geographically concentrated demand (e.g., 15 states account for 80% of U.S. load).
AI’s Impact: AI, especially generative models like ChatGPT, drives exponentially higher energy consumption (10× more per request than standard searches). This trend is expected to amplify with the rise of image, audio, and video generation capabilities.
Future Projections: By 2030, U.S. data center energy use could grow to between 4.6%–9.1% of national electricity consumption, with regional variations (e.g., Virginia could reach 46% under higher growth).
Efficiency and Clean Energy: The report identifies four key strategies for managing growth: improving operational efficiency, enhancing data center-grid collaboration, advancing forecasting models, and adopting energy-efficient hardware, cooling, and clean power sources.
Geographic Implications: Intense load growth in hotspot regions creates local grid challenges. Balancing development with infrastructure timing remains critical.
Energy Distribution: Energy use varies by state (e.g., Texas, Virginia, California), with Texas projected to reach 10.6% under higher growth, scrambling resources due to core load’s increase.

Improve Operational Efficiency: Employ energy-efficient IT hardware, advanced cooling systems (e.g., liquid immersion), and optimize AI algorithms.
Enhance Grid-Center Collaboration: Develop shared energy economy models for better load management and flexibility.
Boost Forecasting: Utilize analytics and modeling to predict localized demand and plan grid investments.
Drive Clean Energy Use: Integrate renewable sources and explore nuclear, hydrogen, or ammonia-based backup power to reduce carbon footprints.

Findings highlight risks in current infrastructure, showing grid limitations in meeting explosive AI-driven demand.
Projections emphasize geographic unevenness, calling for state-specific strategies alongside federal policies.
Key actions include increasing server utilization, migrating toward carbon-free electricity procurement, and implementing real-time monitoring tools.

展开完整摘要