Cheaper LLM tokens led to bigger AI bills (Jevons paradox)
As AI token prices have dropped dramatically, overall AI spending has increased due to higher usage volumes—a real-world example of Jevons paradox, where cheaper resources lead to greater total consumption.
Background
- Jevons paradox: an economic observation that when a resource becomes cheaper and more efficient to use, total consumption of it often rises rather than falls — because lower cost encourages broader use.
- LLM tokens are the basic units text-based AI models process (roughly ¾ of a word each). Companies like OpenAI, Anthropic, Google, and Meta have been dramatically cutting per-token prices, partly through competition and partly through model improvements (e.g. GPT-4o, Claude 3.5, Gemini).
- The article argues that cheaper AI inference is triggering Jevons paradox: developers and businesses ramp up usage so much (more queries, bigger contexts, agent loops) that their total monthly AI spend increases, even though each token costs less.