Stanford’s 2026 AI Index reveals sweeping changes in research, economy, and public trust

Stanford University’s Human-Centered Artificial Intelligence institute has released its 2026 AI Index Report, one of the most comprehensive annual assessments of artificial intelligence. The report draws on data across research, economics, policy, and public opinion to provide a detailed picture of where AI stands today.

Capability is advancing faster than the tools to measure it

AI models are improving at a pace that is outrunning the benchmarks designed to track that progress. On Humanity’s Last Exam, a test built to favor human experts, frontier models gained 30 percentage points in a single year. Evaluations that were expected to remain challenging for years are being saturated within months.

At the same time, the report highlights what researchers call jagged intelligence. Google DeepMind’s Gemini Deep Think scored enough points to earn a gold medal at the 2025 International Mathematical Olympiad. Yet on ClockBench, which tests the ability to read analog clocks, the top model answered correctly only 50.1% of the time, compared with 90.1% for humans. High performance in one area does not predict reliable performance in another.

In professional domains such as tax, mortgage processing, corporate finance, and legal reasoning, models are scoring between 60% and 90% on structured evaluations. However, the top 15 models are often separated by just 3 percentage points, and the report notes that high reliability in these fields remains a significant challenge.

Who is building AI, and at what cost

Industry produced more than 90% of notable AI models in 2025. At the same time, the most capable systems have become the least transparent. Training code, parameter counts, dataset sizes, and training duration are no longer disclosed for several of the most resource-intensive models, including those from OpenAI, Anthropic, and Google.

The environmental cost of this development is growing. The report estimates that training Grok 4 produced approximately 72,816 tons of CO2 equivalent. AI data center power capacity has reached 29.6 gigawatts, comparable to New York state at peak demand. More useful comparisons, e.g. to other Internet and cloud services, are unfortunately missing.

Global AI compute capacity has grown 3.3 times per year since 2022. A single company, TSMC in Taiwan, fabricates nearly every leading AI chip, making the global hardware supply chain dependent on one foundry. The United States hosts more than 5,400 AI data centers, more than ten times any other country.

Investment surges while the workforce picture shifts

Global corporate AI investment more than doubled in 2025. Private investment grew by 127.5% and now accounts for 60% of the total. The United States committed 23 times more in private AI investment than China, though the report notes that Chinese government guidance funds have deployed an estimated $184 billion into AI firms since 2000.

Despite this investment, AI adoption among consumers in the United States lags behind many other countries. The U.S. ranks 24th globally in generative AI adoption at 28.3%, while Singapore reaches 61% and the United Arab Emirates 54%.

The labor market effects of AI are beginning to appear, though unevenly. Employment among software developers aged 22 to 25 has fallen nearly 20% from 2024. One-third of organizations surveyed expect AI to reduce their workforce in the coming year, though large-scale job losses have not yet appeared in overall employment figures.

Public opinion: optimism and anxiety are rising together

Globally, 59% of respondents say AI products and services offer more benefits than drawbacks, up from 55% the previous year. At the same time, 52% say these products make them nervous. Both numbers are increasing simultaneously.

Trust in government to regulate AI responsibly varies sharply by country. In the United States, only 31% of respondents said they trust their government to do so, the lowest figure of any country surveyed. The global average was 54%. Across all 50 U.S. states, concern about too little AI regulation outweighs concern about too much.

Globally, the European Union is trusted more than the United States or China to regulate AI effectively, with a median of 53% expressing trust in the EU across 25 countries surveyed by Pew Research.