How Partnership Structure Shapes AI Patent Performance: Evidence from US and Chinese Digital Platform Ecosystems

Collaborative innovation has become a defining feature of AI-intensive industries, yet the conditions under which formal co-ownership arrangements yield superior technological outcomes remain empirically underexplored. This study examines how the architecture of co-patenting configurations relates to the citation-based performance of AI inventions developed by leading digital platform firms. Drawing on collaborative governance theory and national institutional logics, we analyze 26,488 AI patent families filed between 1995 and 2022, retrieved from Questel Orbit and ORBIS IP databases through a hybrid IPC-keyword classification scheme. Negative binomial regression models are estimated separately for US (N=16,939) and Chinese (N=9,549) sub-samples to capture structural heterogeneity across institutional contexts. Findings reveal that in the US, formal co-ownership neither enhances nor diminishes citation outcomes once baseline patent characteristics are controlled for. In China, inter-firm co-patenting is negatively associated with technological impact, whereas public-private partnerships involving universities and research institutes generate markedly higher forward citations. These results qualify the assumption that openness invariably amplifies innovation performance, demonstrating instead that governance logic and institutional complementarity are decisive determinants of collaborative impact.