The State of AI in 2026: Stanford's Annual AI Index Reveals a Technology Sprinting Faster Than the World Can Manage
AI / AI Tech Trends | 6 min read
Stanford University's Institute for Human-Centered Artificial Intelligence has released the 2026 AI Index — the annual comprehensive assessment of artificial intelligence's progress, adoption, economic impact, workforce effects, energy footprint, and regulatory landscape. The findings present a technology advancing faster than the systems designed to measure, govern, or adapt to it. Despite predictions of a plateau, top AI models keep improving. People are adopting AI faster than they adopted the personal computer or the internet.
"I am stunned that this technology continues to improve, and it's just not plateauing in any way."
— Yolanda Gil, Computer Scientist, University of Southern California, co-author of the 2026 AI Index
US vs. China: A Race Tightening to Razor-Thin Margins
According to Arena — a community-driven ranking platform — the US and China are now almost neck and neck on AI model performance. As of March 2026, Anthropic leads, trailed closely by xAI, Google, and OpenAI, with Chinese models (DeepSeek and Alibaba) lagging only modestly. Competition has shifted to cost, reliability, and real-world usefulness.
Arena AI Model Rankings — March 2026
The US and China have complementary AI strengths: while the US has more powerful models, more capital, and an estimated 5,427 data centres (more than 10 times any other country), China leads in AI research publications, patents, and robotics. Transparency is also declining — companies no longer disclose training code, parameter counts, or dataset sizes, making independent safety research harder.
Jagged Intelligence: Where AI Excels and Where It Fails
AI models now meet or exceed human expert performance on several PhD-level benchmarks. SWE-bench Verified scores jumped from ~60% in 2024 to almost 100% in 2025. Yet robots succeed in only 12% of household tasks, and a popular maths benchmark has a 42% error rate — signs of AI's "jagged intelligence."
AI vs. Human performance (% score)
AI Is Starting to Affect Jobs
Within three years of going mainstream, AI is now used by more than half of people worldwide — faster than the PC or the internet. An estimated 88% of organisations use AI. A 2025 Stanford study found employment for software developers aged 22–25 fell nearly 20% since 2022. A McKinsey 2025 survey found a third of organisations expect AI to shrink their workforce. AI boosts customer service productivity by 14% and software development by 26%.
Early-career employment change since 2022 — age 22–25
age 22–25
age 22–25
age 22–25
Source: Stanford Digital Economy Lab, 2025. Bars show magnitude of decline from 2022 baseline.
People Have Complicated Feelings About AI
Globally, 59% of people think AI will provide more benefits than drawbacks, while 52% say it makes them nervous. The biggest expert-public gap is around work: 73% of AI experts think AI will positively impact how people do their jobs — only 23% of the American public agrees.
% expecting positive AI impact in next 20 years (Pew survey)
Governments Are Struggling to Regulate AI
The EU AI Act's first prohibitions took effect. US states passed a record 150 AI-related bills in 2025, while the US federal government moved toward deregulation. California enacted SB 53 and New York passed the RAISE Act. As Gil noted: "Governments are cautious to regulate AI because we don't understand many things very well."
AI-related bills passed by US states, 2016–2025
Source: Stanford 2026 AI Index. Record 150 bills passed by US states in 2025.
The Cost of Speed: Energy and Supply Chain Fragility
AI data centres worldwide can now draw 29.6 gigawatts of power — enough to run the entire state of New York at peak demand. Annual water use from running GPT-4o alone may exceed the drinking water needs of 12 million people. One company in Taiwan — TSMC — fabricates almost every leading AI chip, creating a concentrated vulnerability in the global AI infrastructure stack.
Key Takeaways
- •Stanford's 2026 AI Index presents AI as advancing faster than the systems designed to measure, govern, or adapt to it. Top models continue improving despite plateau predictions; people are adopting AI faster than any prior technology; AI companies are generating revenue faster than any previous tech boom.
- •US-China AI race is near neck-and-neck — Anthropic leads, followed by xAI, Google, OpenAI, then DeepSeek and Alibaba by small margins. US: more powerful models, capital, 5,427 data centres (10x any other country). China: research publications, patents, robotics leadership.
- •"Jagged intelligence": AI meets or exceeds human performance on PhD-level benchmarks and SWE-bench (near 100% in 2025), but robots succeed in only 12% of household tasks. Benchmarks are broken — a popular maths benchmark has a 42% error rate.
- •Jobs: 88% of organisations use AI; early-career software developer employment down nearly 20% since 2022; a third of organisations expect AI-driven workforce shrinkage. AI productivity: 14% gain in customer service, 26% in software development. Energy: 29.6GW global draw; GPT-4o water use may exceed drinking needs of 12 million people.
- •Expert-public trust gap: 73% of AI experts expect positive job impact from AI vs. 23% of US public. Regulation: EU AI Act first prohibitions took effect; US states passed record 150 AI bills in 2025; California SB 53 and New York RAISE Act enacted; TSMC fabricates almost every leading AI chip — a fragile single point in the global supply chain.