Abstract: Based on the knowledge spillover theory, this study constructs a unique database measuring the development level of artificial intelligence (AI) basic research with the number of published deep learning papers. Combining data from listed companies between 2007 and 2019, it examines the impact of urban AI basic research levels on local enterprises’ innovation performance. The empirical results show that the development of urban AI basic research significantly promotes local enterprise innovation, and this effect becomes more pronounced as the industry relies more on the adoption of AI technology. The conclusion remains robust after addressing endogeneity with the instrumental variable approach and undergoing a series of robustness tests. Mechanism analysis reveals that AI basic research enhances enterprise innovation by improving R&D efficiency, increasing the degree of R&D capitalization, and optimizing the human capital structure. Heterogeneity analysis further indicates that the innovation-promoting effect is more significant in cities with a larger number of AI startups, stronger intellectual property protection, and better broadband network infrastructure. Additionally, the strengthening of AI basic research significantly boosts innovation performance in both state-owned and non-state-owned enterprises. Economic management departments and enterprises should attach great importance to the empowering effect of basic research in the AI era, particularly AI basic research represented by deep learning. They should vigorously promote the deep integration of industry-university-research, accelerate the industrialization of AI basic research achievements, and particularly advance the transition of key technologies centered on deep learning from theoretical breakthroughs to practical applications. Attention should also be paid to the heterogeneity of AI basic research’s impact, and digital transformation and innovation development should be advanced based on industry, regional, and enterprise characteristics.

Key words: artificial intelligence basic research, enterprise innovation performance, knowledge spillover, industry-university-research integration