Modeling Early Diffusion of AI Patents: Institutional Roles, Compute Intensity, and Knowledge Diversity

HRMARS - Purpose: This study examines the factors that accelerate early citations in artificial intelligence (AI) patents and investigates how these determinants vary across countries. By integrating perspectives from innovation management and engineering management, the research aims to clarify how collaboration structures, technological attributes, and knowledge diversity influence the diffusion speed of AI-related inventions. Design/methodology/approach: This study utilizes a global patent dataset encompassing major countries to estimate the time to first citation for each AI patent using Cox proportional hazards models. We test five hypotheses: university collaboration, government research institution involvement, computational resource coherence, technological prominence, and knowledge diversity. Findings: The results above reveal that the initial citation speed of AI patents is explained by the development of AI computational resources (H3), the visibility of the technology (H4), the diversification of application domains (H5), and the nature of the research entities (H1, H2). Research limitations/implications: This study focuses on patent-level data and does not include other R&D outputs such as research papers. It also excludes firm-level strategic variables, such as R&D intensity. Future research could deepen causal inference by integrating firm characteristics and longitudinal collaborative networks. Practical implications: These findings provide practical insights for R&D managers and policymakers. Strengthening industry-academia-government collaboration and expanding R&D investment may accelerate the adoption of AI technologies. Originality/value: This study is among the first to apply survival analysis to AI patents across multiple countries, combining claims-level text analytics with innovation management theory. It contributes to understanding how engineering and organizational factors jointly shape the diffusion dynamics of AI innovations.