While artificial intelligence (AI) partnerships between multinational technology firms and local companies offer significant value creation potential, we lack theoretical understanding of how these relationships evolve across different stages of AI capability development. Prior co-opetition research has predominantly focused on static partnership arrangements or linear learning trajectories, overlooking the unique dynamics created by AI's progression through distinct intelligence levels. Drawing on a 10-year longitudinal study of SenseTime's partnerships and formal modeling, we develop a novel theoretical framework explaining how the evolution through mechanical, analytical, intuitive, and empathetic AI capabilities creates distinct partnership dynamics requiring different balancing mechanisms between knowledge sharing and protection. Our findings advance co-opetition theory by revealing how environmental uncertainty paradoxically enables more sustainable partnership structures by increasing the value of flexible arrangements. We contribute to international business theory by identifying specific mechanisms that enable sustained value creation in cross-border AI partnerships while maintaining strategic flexibility.
