Each year, Stanford’s Institute for Human-Centered Artificial Intelligence (HAI) releases its AI Index report. The study offers a sweeping, data-driven survey of where artificial intelligence stands. The 2026 edition is the most revealing yet: a technology leaping ahead on performance metrics, reshaping economies and classrooms, and leaving regulators scrambling to catch up.

Here are the ten findings that matter most:

Performance is accelerating, not plateauing

The narrative of diminishing returns in AI development is not holding up. Industry produced over 90% of notable frontier models in 2025, and several now meet or exceed human baselines on PhD-level science questions, multimodal reasoning, and competition mathematics. On one key coding benchmark — SWE-bench Verified — performance jumped from 60% to near 100% in a single year. Organizational adoption reached 88%, and four in five university students now use generative AI regularly.

“AI capability is not plateauing. It is accelerating and reaching more people than ever.”

The U.S.–China gap has effectively closed

What was once a clear American-led AI model performance is now a competitive race. U.S. and Chinese models have traded the top spot multiple times since early 2025. DeepSeek-R1 briefly matched the leading U.S. model in February 2025, and as of March 2026, Anthropic’s top model leads by just 2.7%. The U.S. still produces more top-tier models and higher-impact patents; China leads in publication volume, patent output, and industrial robot installations. South Korea stands out for innovation density, leading the world in AI patents per capita.

The hardware supply chain runs through one building in Taiwan

The U.S. hosts more than 5,400 data centers, dwarfing every other country. But almost every leading AI chip is fabricated by a single company — TSMC — at a single location in Taiwan. That concentration is a strategic vulnerability that policymakers and executives are only beginning to reckon with. A TSMC facility in the United States began operations in 2025, but the dependency remains acute.

Rendering of the Taiwan Semiconductor Factory, presumably one of the five coming to the United States. The Arizona facility has already sold out production through 2027. The Arizona operation employees approximatley 3000. https://www.tsmc.com/static/abouttsmcaz/index.htm

AI has a jagged frontier

One of the report’s most striking — and humbling — findings: the same models winning gold medals at the International Mathematical Olympiad read analog clocks correctly just 50.1% of the time. AI agents went from 12% to roughly 66% task success on OSWorld (a benchmark testing real computer tasks), but still fail about one in three attempts on structured tests. The frontier of capability is not a smooth curve; it is jagged and unpredictable.

Responsible AI is falling behind

Nearly all leading AI developers report results on capability benchmarks. Reporting on responsible AI benchmarks — safety, fairness, transparency — remains inconsistent. Documented AI incidents rose to 362 in 2025, up from 233 the year before. Complicating matters further, recent research found that improving one responsible AI dimension, such as safety, can degrade another, such as accuracy. The tradeoffs are real and poorly understood.

America leads in investment but is losing its talent advantage

U.S. private AI investment reached $285.9 billion in 2025 — more than 23 times China’s $12.4 billion, though China’s state-guided funds make direct comparisons imprecise. The U.S. also led in new AI company formation, with 1,953 freshly funded startups. Yet the number of AI researchers and developers choosing to relocate to the United States has dropped 89% since 2017, with an 80% decline in the last year alone. That is a canary worth watching.

Adoption is outpacing the internet or the PC

Generative AI reached 53% population adoption within three years — faster than the personal computer or the internet. That said, the pace varies sharply by country and correlates strongly with GDP per capita. Singapore (61%) and the UAE (54%) show higher-than-expected adoption. The U.S. ranks 24th globally at 28.3%. The estimated annual value of generative AI tools to U.S. consumers reached $172 billion by early 2026, with the median value per user tripling between 2025 and 2026 — much of it captured for free.

Education is still catching up

Over 80% of U.S. high school and college students now use AI for school-related work, but only half of middle and high schools have any AI policy in place, and just 6% of teachers describe those policies as clear. Outside the classroom, AI engineering skills are growing fastest in the UAE, Chile, and South Africa. New AI PhDs in the U.S. and Canada increased 22% from 2022 to 2024 — but the newly minted PhDs driving that growth are heading into academia, not industry. Full Education Report.

AI sovereignty is reshaping geopolitics

National AI strategies are multiplying, particularly among developing economies, and state-backed investments in supercomputing infrastructure are rising. Yet model production remains concentrated in the U.S. and China. Open-source development is beginning to redistribute participation: contributions from the rest of the world now outpace Europe on GitHub, fueling more linguistically diverse models and benchmarks. The question of who controls the infrastructure of intelligence is becoming a defining feature of international politics.

Experts and the public are living in different realities

Perhaps the most striking finding in the report is the perception gap between AI experts and the general public. When asked about AI’s impact on how people do their jobs, 73% of experts expect a positive effect — compared with just 23% of the public. That 50-point gap recurs across questions about the economy and medical care. On trust in governments to regulate AI, the findings are equally striking: the United States reported the lowest level of trust in its own government among surveyed countries, at 31%. Globally, the EU is trusted more than either the U.S. or China to regulate AI effectively.

“Among surveyed countries, the United States reported the lowest level of trust in its own government to regulate AI.”

The 2026 AI Index is a benchmark document — not optimistic, not pessimistic, but rigorously empirical about a technology that is moving faster than almost anyone predicted. Read alongside the headlines, it offers something rare: data over narrative. For anyone trying to understand where AI is actually going, it is essential reading.