When science meets geopolitics: global AI research network transformation (2000–2025)

Abstract Global AI research has shifted from an open, collaborative enterprise to a domain shaped by strategic rivalry. This study examines the evolution of international AI research collaboration networks from 2000 to 2025 through bibliometric and social network analysis of 1.4 million publications from the Web of Science. We identify four phases: a unipolar Western-led network (2000–2009), China’s rise and bipolar emergence (2010–2016), peak collaboration amid geopolitical strain (2017–2021), and strategic bifurcation (2022–2025). The findings reveal asymmetric centrality: China dominates in publication volume, while the US retains structural influence. Rather than emerging solely from intrinsic scientific dynamics, collaboration patterns are increasingly shaped by political interventions and techno-industrial strategies. Technological breakthroughs catalyse investment cycles that reshape global networks, while geopolitical forces disrupt established partnerships. The study highlights the policy imperative to support multi-polar research ecosystems, empower intermediary ‘bridge’ nations, and pursue nuanced strategies that balance competition with sustained international cooperation.