From infrastructure to applications: how foundational, core and general-purpose AI technologies jointly shape sustainable urban innovation in China
Xupei Wang, Yi Zhang, Minglian LongPurpose
This study aims to investigate how different layers of artificial intelligence (AI) technologies jointly shape urban innovation, addressing the puzzle of why cities with high AI adoption often display uneven innovation outcomes. Rather than treating AI as a single technology, the study conceptualises AI as a three-tier capability structure consisting of foundational, core and general-purpose technologies, and examines whether urban innovation benefits more from isolated technological investments or from coordinated development across these layers in an emerging-economy context.
Design/methodology/approach
Using patent-based indicators, this study constructs city-level measures of foundational, core and general-purpose AI technologies for a balanced panel of 120 Chinese cities from 2010 to 2021. Fixed-effects models with interaction terms are used to estimate both direct and complementary effects across AI layers. An instrumental-variable strategy based on historical communication infrastructure is further applied to address potential endogeneity concerns, and heterogeneity analyses are conducted across different urban contexts.
Findings
The results show that the innovation effects of AI are layered and uneven. Foundational and core AI technologies are more strongly associated with urban innovation capacity, whereas the contribution of general-purpose AI technologies is more contingent on local development conditions. The findings further suggest that complementarities across AI layers are conditional and that sustainable urban innovation depends on both upstream capability accumulation and downstream application diffusion.
Research limitations/implications
The analysis is based on city-level data from 2010 to 2021 and therefore does not fully capture the post-2022 diffusion of generative AI. Future research could extend the framework to examine whether large-model technologies reshape the relationships among different AI layers and urban innovation outcomes.
Practical implications
The study suggests that policymakers should adopt differentiated AI development strategies, balancing application expansion with sustained investment in foundational infrastructure, core technologies and local innovation capabilities.
Originality/value
This study advances innovation research by reconceptualising AI as a layered technology stack rather than a monolithic general-purpose technology. It reveals a hierarchical complementarity mechanism through which AI capabilities translate into urban innovation, offering new insights into why application-led AI strategies often fail. The findings provide actionable implications for urban and regional innovation policy in emerging economies by highlighting the importance of balanced, context-sensitive AI capability development.