AI Insights: Key Findings from Stanford’s 2025 AI Index Report

On April 7, 2025, the Stanford Institute for Human-Centered Artificial Intelligence (HAI) unveiled the 2025 AI Index Report, providing a comprehensive analysis of the global development of artificial intelligence. This report builds on several previous years of research, with significant changes observed since its first benchmark report in 2022.

The report is rich with data, highlighting a number of crucial findings for enterprise technology leaders:

• The United States produced 40 notable AI models in 2024, far exceeding China’s 15 models and Europe’s 3.
• Training compute for AI models has doubled approximately every five months, while dataset sizes increase every eight months.
• The costs associated with AI model inference have plummeted, experiencing a staggering 280-fold reduction from 2022 to 2024.
• Global private investments in AI reached $252.3 billion in 2024, marking a 26% increase from the previous year.
• A significant 78% of organizations now report utilizing AI, an increase from 55% in 2023.

For enterprise IT leaders, the report outlines critical insights concerning model performance, investment trends, implementation challenges, and competitive dynamics shaping the technology landscape. Here are five essential takeaways from the AI Index:

1. The democratization of AI power is accelerating

A particularly notable finding is the rapid decrease in costs associated with high-quality AI, making it far more accessible. This shift marks a significant change from the findings of the 2024 Stanford report. Nestor Maslej, research manager for the AI Index at HAI, stated, ‘I was struck by how much AI models have become cheaper, more open, and accessible over the past year.’ Although training costs are still substantial, the expense of developing high-quality non-frontier models has dramatically decreased.

The report provides startling numbers: the inference cost for an AI model performing at GPT-3.5 levels dropped from $20.00 per million tokens in November 2022 to just $0.07 per million tokens by October 2024—an extraordinary 280-fold reduction in just 18 months.

Further narrowing performance gaps between closed and open-weight models has also been observed; the gap between leading closed models, such as GPT-4, and top open models, such as Llama, decreased from 8% in January 2024 to just 1.7% by February 2025.

IT leader action item: Organizations should reassess their AI procurement strategies, as they now have more options in open-weight models or significantly cheaper commercial APIs.

2. The gap between AI adoption and value realization remains substantial

Despite a rise in AI adoption—with 78% of organizations now using AI in at least one business function—the tangible impact of AI on business performance remains limited. Maslej noted, ‘We have limited data on what separates organizations that achieve massive returns to scale with AI from those that do not.’

The report indicates that while some companies report modest financial improvements from generative AI, genuine benefits are still elusive, with only 47% of businesses using generative AI in strategy and corporate finance reporting revenue increases typically below 5%.

IT leader action item: Focus on measurable use cases that promise clear ROI and develop robust AI governance and measurement frameworks.

3. Specific business functions show stronger financial returns from AI

Insights into which functions benefit most financially from AI usage reveal significant disparities. ‘On the cost side, AI appears to benefit supply chain and service operations functions the most,’ said Maslej. Notably, 61% of organizations using generative AI in supply chain and inventory management report cost savings.

Furthermore, 70% of organizations using AI for strategy and corporate finance have seen revenue increases. Service operations and marketing/sales functions also demonstrate substantial value potential.

IT leader action item: Prioritize investments in functions where financial returns are most pronounced, such as supply chain optimization and strategic planning.

4. AI shows strong potential to equalize workforce performance

The report outlines that AI tools can enhance productivity for lower-skilled workers more significantly than for their higher-skilled counterparts, illustrating AI’s ability to level the playing field. For example, low-skilled workers in customer support reported a 34% productivity boost with AI assistance, while high-skilled workers showed minimal gains.

Maslej noted, ‘These studies indicate that AI has strong positive impacts on productivity and tends to benefit lower-skilled workers more than higher-skilled ones.’

IT leader action item: Consider deploying AI as a strategy to develop workforce skills and enhance team performance.

5. Responsible AI implementation remains an aspiration, not a reality

Awareness of AI risks is growing, but substantial gaps remain between risk recognition and mitigation efforts. For instance, while 66% of organizations acknowledge cybersecurity as an AI risk, only 55% actively mitigate it. The absence of sound management practices could have serious consequences for organizations.

IT leader action item: Implement sound governance practices for responsible AI as technical capabilities advance rapidly.

Looking Ahead

The Stanford AI Index Report paints a promising picture of increasingly advanced and accessible AI technology, yet organizations must grapple with fully realizing its potential. IT leaders are urged to adopt focused and strategic AI implementations with demonstrable ROI while emphasizing accountability and workforce enhancement.

As Maslej observed, ‘This shift points toward greater accessibility and, I believe, suggests a wave of broader AI adoption may be on the horizon.’

• 0 Comments
• Stanford Report