We fine-tune a suite of Large Language Models (LLMs) to identify AI patents with high out-of-sample precision and recall and study AI inventor migration between the US and China. We document a sharp increase in both volume and quality of the US-to-China migration, particularly after the onset of US–China trade tensions in 2018. These migrants significantly increase their patent output and exploratory innovation, with the largest gains observed among those joining research institutions, having greater experience, and a consistent research focus. In contrast, China-to-US migrants show declined patenting activity but gains in originality. Post-migration, immigrants significantly expand their collaborator networks. AI inventor migration generates significant and persistent local spillovers through multiple layers of collaborators, benefiting both countries. These findings highlight China’s growing attractiveness to top-tier AI talent and suggest that recent US policy changes may have accelerated the global reallocation of innovation activity.