AI's impact on energy efficiency and electricity consumption: Evidence from AI green computing power in China's National Green Data Centers

: This study pioneers the analysis of the impact of AI green computing power on urban electricity utilization efficiency and consumption. Based on 4,519 observations from 279 prefecture-level cities in China between 2004 and 2022, we found that AI green computing capabilities can reduce electricity energy intensity by 20%, improve electricity energy efficiency, reduce electricity consumption, and thus mitigate the energy rebound effect. Efficiency gains are more pronounced in non-coal resource-based cities, non-old industrial base cities, non-environmental protection cities, and cities with higher levels of data elementization. Consumption reductions are more significant in coal-reliant, non-old industrial base, and low data elementization cities. Additionally, further investigation reveals that the scale effects of green data center construction exert a more pronounced influence on electricity utilization efficiency; however, at the current stage, these effects are particularly prominent only in the telecommunications, internet, and public institutions sectors. Finally, analyses of the new energy supply side and carbon emission cost constraints on the demand side demonstrate that the synergistic effects of these dual policies significantly reinforce the role of green computing power in green data centers in elevating electricity utilization efficiency. These findings offer valuable insights for policies that balance AI deployment with energy and environmental sustainability.