Industry players now dominate the development of cutting-edge artificial intelligence, accounting for nearly 90% of notable models released in 2024, up sharply from 60% the year prior, according to Stanford University’s 2025 AI Index report. The shift underscores a growing consolidation of AI innovation within the private sector, even as top-tier academic research remains influential in citation metrics.
The report highlights a steep acceleration in model scale and development intensity. Training compute requirements are doubling every five months, datasets every eight, and energy consumption continues to rise on an annual basis. This rapid growth reflects the intensifying demands of building competitive frontier models, systems at the leading edge of capability and scale.
But as investment and infrastructure scale up, the performance gap between leading models is shrinking. In 2023, the score difference between the top-ranked model and the 10th stood at 11.9%. By 2024, that gap had narrowed to just 5.4%. The two best-performing models are now separated by a margin of only 0.7%, suggesting a maturing field with diminishing returns at the high end.
The findings reflect a market that is both highly competitive and increasingly saturated at the top. While scale remains a differentiator, the pace of improvement is compressing, leaving less room for breakthrough leaps and more pressure to optimize on margins.
Regarded as one of the most authoritative sources for tracking global AI progress, Stanford’s latest data suggests that while industry-led innovation continues to push the frontier forward, the space is becoming more contested, and more resource-intensive, than ever before.
