China has "nearly erased" America’s lead in AI—and the flow of tech experts moving to the U.S. is slowing to a trickle, Stanford report says
The U.S. lead in AI bot performance shrank from 300+ points to just 39 in under three years.
A new report from Stanford's Institute for Human-Centered Artificial Intelligence (HAI) delivers a stark assessment of the shifting global AI landscape. The 2026 AI Index finds that China has "nearly erased" the United States' longstanding lead in artificial intelligence. The most telling metric is the dramatic narrowing in AI model performance. In May 2023, the top U.S. model, OpenAI's GPT-4, led China's best by over 300 Arena points (a key benchmark for large language model performance). By March 2026, that gap had collapsed to just 39 points, with Anthropic's Claude Opus 4.6 leading China's Dola-Seed 2.0 by a mere 2.7%.
Beyond raw model scores, China continues to dominate in foundational metrics like the volume of AI patents, academic publications, and the industrial rollout of robots. Compounding this shift is a critical slowdown in the brain drain that once fueled U.S. innovation: the flow of Chinese AI experts and tech talent to America has dwindled to a trickle. This combination of rapid technical catch-up and retained domestic talent creates a new paradigm of near-parity between the two superpowers, moving the competition from a race for leadership to a battle for incremental advantage across different sectors.
- The performance gap between top U.S. and Chinese AI models shrank from over 300 Arena points in 2023 to just 39 points in 2026.
- China maintains global leads in AI patents, research publications, and deployment of industrial robots.
- The migration of Chinese AI experts to the U.S. has slowed dramatically, reducing a key historical advantage.
Why It Matters
This signals a new era of intense, head-to-head competition in AI, impacting global tech strategy, investment, and policy.