Stanford's 2026 AI Index Report reveals the performance gap between top US and Chinese AI models has collapsed to just 2.7%, down from 17.5–31.6 percentage points in May 2023, despite the US spending 23x more on private AI investment ($285.9B vs $12.4B). China leads in AI patents (69.7% of global filings), publications (23.2%), and industrial robot installations (9x the US rate). AI talent migration to the US has dropped 89% since 2017, with 80% of that decline in the past year alone. On capabilities, models now score near 100% on SWE-bench and 93% on graduate-level science, but real-world reliability lags. Generative AI reached 53% population adoption in three years, yet only 31% of Americans trust their government to regulate it. The open-vs-closed source gap widened to 3.3%, favoring US proprietary labs. Environmental costs are rising sharply, with Grok 4 training emitting 72,816 tonnes of CO2 equivalent.
Table of contents
Where each country leadsThe talent crisisWhat AI can and cannot doAdoption, trust, and regulationThe environmental costWhat the numbers meanSort: