When and how artificial intelligence influences the coupled coordination of energy security, energy equity, and environmental sustainability in China

Energy security, energy equity, and environmental sustainability constitute the three core objectives of energy policy, yet they have long faced the persistent challenge of the energy trilemma, which makes their coordinated advancement difficult. In the context of the new wave of technological revolution, artificial intelligence (AI), as a general-purpose technology, may provide a potential pathway to alleviate this trilemma. Based on panel data of Chinese cities, this study constructs a theoretical model and an empirical framework to systematically examine the impact of AI on the coupled coordination degree of the energy triangle (ETCCD) and its underlying mechanisms. The findings include four aspects. First, theoretical derivations suggest that the effect of AI on ETCCD is uncertain, primarily depending on whether the efficiency gains generated by technological improvements can sufficiently offset the additional energy consumption incurred by AI systems themselves. When the marginal benefits from technological improvement substantially exceed the incremental energy costs, AI exhibits a positive effect on ETCCD; otherwise, it may produce adverse outcomes. Subsequent empirical evidence quantitatively supports the positive impact of AI, showing that AI adoption increases ETCCD by 0.45 percentage points, with the result being robust across multiple sensitivity checks. Second, analysis from technological, organizational, and environmental dimensions indicates that AI promotes the coordinated development of energy security, equity, and sustainability by driving green technological innovation, optimizing human capital allocation, and mitigating climate risks. Third, the energy performance of AI demonstrates significant heterogeneity. The positive effect of AI on ETCCD is more pronounced in regions southeast of the Hu Huanyong Line, in cities with higher future industrial levels, and in cities with shorter distances to optical fiber nodes. Furthermore, computational power positively moderates this relationship by enhancing AI's real-time optimization capability.