Artificial intelligence (AI) can enhance efficiency and reduce emissions, yet also intensifies energy demand through computation and infrastructure. This paradox raises a key question: whether AI policy instrument can reduce urban carbon emission intensity (CI)? Using the quasi-natural experiment of "National New Generation Artificial Intelligence Innovation Development Pilot Zone" (AIPZ), this paper investigates the policy's local and spillover impacts on urban CI. Results show that AIPZ reduces CI in pilot cities and neighboring regions. Mechanism analysis reveals that the improvements in labor productivity and green technology innovation serve as main mitigation pathways across local and neighboring areas, while rising computing power demand increases emissions. Energy efficiency gains, though effective in reducing local CI, generate no significant spatial spillovers. Together, the mitigation effects dominate, yielding a net carbon-mitigating outcome especially in eastern, resource-based, and high digital-infrastructure cities. These findings offer insights for fostering AI development via coordinated regional strategies.
Yu et al. (Mon,) studied this question.