At present, the development logic of the supply chain has shifted from “efficiency driven” to “resilience driven”. How to effectively improve the AI level of enterprises along the supply chain and narrow the technological gap across chain positions has become the key to facilitating supply chain adaptation to the new landscape and enhancing overall collaboration efficiency. This paper uses supply chain linkage data of A-share listed enterprises from 2012 to 2024 to examine the bidirectional spillover effect, mechanism, boundary conditions and economic consequences of artificial intelligence (AI) adopted by core enterprises from the supply chain perspective. The results show that AI of core enterprises exerts a significant bidirectional spillover effect along the supply chain, yet this effect presents obvious asymmetry across supply chain positions: the diffusion toward upstream suppliers is stronger than that toward downstream customers. Mechanism tests reveal that the evolution of supply chain organizational structure serves as an important foundation for cross-enterprise AI diffusion. Supply chain diversification, specialization and co-creation facilitate the bidirectional transmission of AI within supply chains. Heterogeneity analysis indicates that the bargaining power of core enterprises, inter-enterprise technological gaps, spatial frictions and knowledge absorption capacity jointly shape the boundary conditions for AI diffusion in supply chains. Extended analyses demonstrate that AI spillovers in supply chains can effectively mitigate supply chain risks for upstream and downstream enterprises and exert positive impacts on improving the collaborative stability of supply chains. This paper provides a novel analytical framework for understanding the co-evolution of AI within supply chain systems, and offers empirical evidence and policy implications for advancing the intelligent transformation of supply chains and the construction of modern industrial systems.