Stanford University's Institute for Human-Centered Artificial Intelligence (HAI) has released its eighth annual AI Index Report, providing a comprehensive analysis of the global artificial intelligence landscape as of May 2025.
The report's most striking finding reveals the democratization of advanced AI access, with inference costs for models performing at GPT-3.5 levels plummeting from $20 per million tokens in November 2022 to just $0.07 by October 2024—a 280-fold reduction. This dramatic cost decrease stems from improvements in hardware performance, energy efficiency, and the emergence of smaller yet highly capable models.
While the United States continues to lead in AI development with 40 notable models produced in 2024 (compared to China's 15 and Europe's 3), the quality gap is rapidly narrowing. On key benchmarks like MMLU and HumanEval, performance differences between top US and Chinese models have shrunk from double-digit percentages in 2023 to near parity by early 2025. This convergence signals an intensifying global AI race with significant geopolitical implications.
The report also highlights concerning trends in AI safety. The AI Incidents Database documented 233 harmful AI-related incidents in 2024—a record high representing a 56.4% increase over the previous year. These incidents included deepfake pornography, false identifications by surveillance systems, and chatbots allegedly encouraging harmful behaviors. Despite these risks, standardized responsible AI evaluations remain rare among major model developers.
Global investment in AI continues to surge, with private funding reaching $252.3 billion in 2024—a 26% year-over-year increase. The US dominates with $109.1 billion invested, nearly 12 times China's $9.3 billion. However, organizational adoption is growing worldwide, with 78% of surveyed companies now using AI in at least one business function, up from 55% in 2023.
The Stanford HAI report, published annually since 2017, has become an authoritative resource regularly cited by governments worldwide as they navigate the complex technical, economic, and ethical dimensions of artificial intelligence development.