The strategic buffer that US policymakers once believed existed between US and Chinese artificial intelligence capabilities has largely evaporated, according to Stanford University’s 2026 AI Index report.
The two countries have traded the lead in model performance several times since the beginning of 2025, with the report clearly describing this gap as effectively closing.
The numbers that will trouble Washington most are the model benchmarks. In February 2025, China’s DeepSeek-R1 briefly matched the best-performing US system.
By March 2026, Anthropic’s leading model had a profit of only 2.7%, a margin small enough to be flipped on the next major release by either side.
The US produced 50 notable AI models by 2025 compared to China’s 30, maintaining a quantitative lead in frontier system development. But the performance gap demonstrated by those additional models has largely disappeared.
China’s superiority is most evident when it comes to intellectual property. Chinese organizations were issued 97,206 patents for AI in 2024, accounting for 74.2% of the 131,121 patents issued worldwide. America’s market share was 42.8% in 2015, which has now come down to 12.1%.
Research output is no different. China was responsible for 17.8% of all AI publications in 2024, compared to just 7.6% for the US. Of all citations to AI research globally, Chinese papers received 20.6%, while the US received 12.6%.
In terms of the 100 most cited AI papers worldwide, China’s presence increased from 33 in 2021 to 41 in 2024, while the US dropped from 64 in 2021 to 46 in 2024.
US private investment in AI is set to reach $285.9 billion in 2025, twenty times more than the amount China has privately invested. In addition, the US launched 1,953 new startups specializing in AI technology in the same year, an order of magnitude more than any other country.
The Stanford University report highlights that China’s figures are underestimated because the government is a major funder of AI technology, and these funds do not go to private markets.
