Half of top AI talent is Chinese
A new survey finds half of the world's top AI talent now comes from China, with nine leading AI universities there and projections that China's talent pool could double the US's by 2028 — a major shift for global recruiting and mobility (chosun.com).
The headline’s numbers are drawn from MacroPolo’s Global AI Talent Tracker, a Paulson Institute project first updated in March 2024 and benchmarked against author affiliations at NeurIPS and similar top conferences. (paulsoninstitute.org) Meta’s push to staff its new “Superintelligence Labs” produced public reporting that the unit grew to roughly 44 people with about half from China, and Bloomberg reported some hires received compensation packages in the hundreds of millions of dollars. (scmp.com) (bloomberg.com) OpenAI’s and DeepMind’s public research-scientist listings specify PhD-level qualifications, a publication record, and the ability to “move easily between theory and code,” while DeepMind job posts explicitly list proficiency in Python and frameworks such as JAX, PyTorch or TensorFlow. (openai.com) (deepmind.google) Hiring-process analyses and mock-interview guides for frontier labs show technical rounds focus on core ML theory and advanced mathematics, noting that candidates with strong engineering skills but “shallow math foundations get filtered out.” (datainterview.com) NeurIPS 2025 affiliation analyses documented a rise in authors who hold dual academic and industry roles and flagged access to large-scale compute as a differentiator for top papers—an ecosystem signal that labs prize candidates who combine rigorous publication records with production-scale model experience. (aiworld.eu) (finance.sina.com.cn) MacroPolo’s tracker and related reporting also recorded declining cross-border mobility of elite researchers since 2019, with more top talent remaining in home institutions or domestic labs—shifting the recruitment dynamic toward competing to retain domestically trained researchers as well as to hire them abroad. (digitalprojectsarchive.org) (paulsoninstitute.org) Profiles of recent hires at OpenAI, DeepMind and Meta show a common pattern: a strong NeurIPS/ICML publication record, PhD-level theoretical credentials, and demonstrable experience shipping or optimizing large models—facts that hiring teams at elite labs list as decisive in selection. (openai.com) (builtin.com) (bloomberg.com)
