Shawn Wang
As AI becomes increasingly integrated into the real sector, identifying how AI development shapes urban carbon emission intensity (CEI) is important for China’s low-carbon transition. Using panel data for 266 prefecture-level (and above) Chinese cities from 2011 to 2020, we examine the effect of AI on CEI and the underlying mechanisms. Four findings emerge. (1) AI is associated with an inverted U-shaped pattern in CEI: CEI rises at early stages of AI development but declines after a turning point, implying an early emission-accelerating effect and a later mitigating (braking) effect. Importantly, by the end of the sample period (2020), AI intensity in most cities still lies on the rising segment of the curve, so the braking effect has not yet become widespread. (2) Mechanism analyses suggest that the nonlinear relationship operates through energy-use intensity, green technological innovation, and industrial structure upgrading. (3) The AI–CEI relationship is heterogeneous across cities by economic development, resource endowments, and urbanization patterns. (4) Spatial models indicate nonlinear spillovers: AI first increases and then decreases CEI in neighboring areas, implying strong regional co-movement. Based on these results, we recommend deepening sectoral AI deployment while promoting a greener AI development pathway, strengthening regional coordination in innovation and coopetition, and adopting place-based policies that reflect local endowments, urbanization forms, and development stages to advance the joint transition toward digital intelligence and low-carbon development.