Stanford's 2026 AI Index provides a comprehensive look at the state of AI across multiple dimensions. Key findings include: US companies lead in notable AI model releases (50 in 2025) while China dominates industrial robotics deployment; global AI compute capacity has grown 3.3x annually since 2022 with Nvidia holding 60%+ market share; AI training emissions are skyrocketing (Grok 4 estimated at 72,000+ tons CO₂); LLMs are rapidly conquering benchmarks including Humanity's Last Exam (now 50%+ accuracy vs 8.8% a year ago); AI investment hit a record $581 billion in 2025, more than double 2024; GitHub AI projects reached 5.58 million; employment impact remains mixed with entry-level jobs declining while senior roles hold steady; and public perception of AI slightly improved globally, though trust in government regulation varies widely by country.

11m read timeFrom spectrum.ieee.org
Post cover image
Table of contents
US companies lead in AI modelsChina leads in roboticsWorld AI compute capacity has grown 3.3x yearly since 2022Training AI models can generate enormous carbon emissionsLLMs are rapidly defeating new benchmarksAI research in medicine sees gainsLLMs still have trouble reading an analog clockAI investment hit a new peak in 2025Software engineers are all-in on AIAI’s overall impact on employment remains unclearOverall public perception of AI (slightly) improvesTrust in AI regulation varies significantly by country

Sort: