The collapse in AI inference costs
The cost of querying an AI model at GPT-3.5-equivalent MMLU performance fell from about USD 20 per million tokens in November 2022 to roughly USD 0.07 by October 2024 — a more than 280-fold reduction over 23 months. This measures inference pricing at a benchmark threshold, not the total cost of industrial AI deployment.
Source: Stanford HAI — AI Index 2025: State of AI in 10 Charts (2025)
What it means
Lower inference prices can improve the economics of processing maintenance text, but this benchmark does not establish the cost or suitability of a specific industrial application. An operator still needs to budget for data preparation, integration, validation, security, monitoring and human review. Image and sensor workloads require their own performance and cost assessment.
Context
The Stanford AI Index tracks the price of achieving a fixed quality threshold (about 64.8% on the MMLU benchmark) rather than the price of a single named model. Depending on the task, the report finds inference prices falling anywhere from 9 to 900 times per year. Because the metric holds quality fixed while hardware and models improve, it captures genuine economic gains rather than simple discounting.
How to interpret this data
About the source: This data comes from Stanford HAI. Use the linked publication to check the observation year, population and definitions behind the chart. Access to a public summary does not imply that the underlying data is freely downloadable or licensed for reuse. Check the publisher's access and reuse terms before redistributing data.
Where this matters: Generative AI in manufacturing, AI agents for industrial maintenance are built on insights like the data shown here. Rather than treat data in isolation, read the deeper guides to see how these trends translate into actionable levers for your plant.
Sector relevance: This dataset is especially relevant to Chemicals, Food Processing. These sectors face the trends and challenges you see in this chart daily — energy cost pressure, the push for decarbonization, adoption of AI and predictive maintenance. Use this data to benchmark your plant against the industry average and identify where you lag or lead.
How to use this data: Take the headline number but look deeper at the chart. Is it growing or shrinking? Which segments or regions drive the trend? Does your plant's data align, or are you an outlier? Outliers are often where the best opportunities hide — either an efficiency gap you can exploit, or a leading practice you can copy.
Related charts
Organisational AI adoption: context for manufacturers
Industrial robot installations worldwide
Robot density in manufacturing
Related topics
Generative AI in Manufacturing: Practical Examples · AI Agents for Industrial Maintenance: What They Are and Where They Help · Large Language Model (LLM) · Generative AI · Machine Learning (Industrial)
Relevant to: Chemicals · Food Processing · Pharmaceuticals