美国电力研究院（EPRI）：2026赋能智能：美国数据中心电力使用与供电策略情景更新白皮书（英文版）（28页）.PDF

1、2026 White PaperPowering Intelligence 2026Updated Scenarios of U.S.Data Center Electricity Use and Power StrategiesAvailable online:powering-02|Powering Intelligence 2026 February 2026TABLE OF CONTENTSEXECUTIVE SUMMARY.3INTRODUCTION.10UNDERSTANDING KEY DATA CENTER POWER METRICS.10Nominal Capacity.11

2、Nameplate Capacity and Power Usage Effectiveness(PUE).11Annual and Peak Electricity Use.11NOMINAL CAPACITY:HISTORY AND FUTURE PROJECTIONS.12Future Scenarios.12ANNUAL AND PEAK ELECTRICITY USE:HISTORY AND FUTURE PROJECTIONS.15SUMMARY OF FUTURE PROJECTIONS.19Comparison to Other Studies and Forecasts.20

3、DATA CENTER LOAD GROWTH IN CONTEXT.22Data Center Shares of Total Electricity Demand.22Comparison to EV Charging Load.24GENERATION AND CAPACITY IMPACTS OF DATA CENTER LOAD.25CONCLUSIONS.27REFERENCES.2803|Powering Intelligence 2026 February 2026Key Messages Data centers,industrial onshoring,and transp

4、ort electrification are driving renewed regional load growth.In the United States and other nations,clusters of new,large loads are testing utilities ability to keep pace and are spurring technical and business innovation.Artificial Intelligence(AI)plays a growing,but uncertain role in data center l

5、oad.AI workloads are estimated to account for 1525%of data cen-ter electricity today and that share is rising rapidly even as non-AI data center demand continues steady growth.However,the pace and scale of AI adoption,the power intensity of AI hardware and algorithms,and power system and supply chai

6、n constraints remain highly uncertain.EPRI projects data centers to consume 9%to 17%of U.S.electricity by 2030,up from 4%to 5%today.EPRI developed Low,Medium,and High growth scenarios of U.S.state-level data center power demand through 2030 using commercial project development data.The Low scenario