Stanford Report Says US-China AI Performance Gap Has Closed

The performance gap between American and Chinese artificial intelligence models has effectively closed, according to the 2026 AI Index report from Stanford University. The two nations have traded the lead multiple times since early 2025, with a leading model from Anthropic holding a narrow 2.7% advantage as of March 2026. Earlier, in February 2025, the DeepSeek-R1 model briefly matched the top-performing American system.

While the United States leads in private investment with $285.9 billion and the production of frontier systems—developing 50 notable models in 2025 compared to China's 30—China has achieved superiority in research volume and intellectual property. In 2024, Chinese organizations accounted for 74.2% of global AI patents and 17.8% of AI publications, while the US share of patents fell to 12.1%. Stanford also noted an 80% drop in global AI researchers moving to the US over the past year.

China's growth is supported by a strong industrial base and government-led data collection. GigaAI executives claim these factors provide a decisive advantage in world models, which simulate 3D environments for robotics and autonomous vehicles. In contrast, US development remains more focused on gaming and AI-assisted design. The report suggests China's total AI spending is likely understated because heavy government funding is not captured in private market data.