AI prices are falling but firms are spending more
The economics of artificial intelligence are shifting as the cost of open-weight models falls, encouraging wider adoption among enterprises. While foundational model inference costs have dropped significantly—with LLM inference prices falling nearly 90% in some cases—companies are increasing overall spending. Rather than relying entirely on a single costly frontier model, firms are using diverse models for different tasks, leading to higher token consumption and more complex agentic workflows. However, this shift raises concerns about measuring true business value versus raw token consumption and increases the computational intensity of AI deployments.
Key Points
- The cost of open-weight models has decreased drastically, making AI adoption more accessible.
- Despite falling token prices, enterprises are spending more on AI due to increased usage and complex multi-model workflows.
- Open-weight models now account for a significant share of token use, challenging proprietary dominance.
- The shift towards agentic workflows and multi-model deployment increases overall computational intensity and spending.
Exam Facts
- LLM inference prices fell by over 90% between November 2022 and October 2024.
- Open-weight models account for about 35% of token use on certain platforms.
Read it. Retain it. Recall it.
Get spaced-repetition flashcards, daily quizzes and offline access — free on Android.