Agentic AI Foundation: API usage up 90% as per-token prices drop

The AI industry is grappling with a cost paradox: while per-token prices have plummeted, overall spending on AI is climbing. According to the Agentic AI Foundation, an open-source standards body hosted by The Linux Foundation and backed by OpenAI, Anthropic, Google, and Microsoft, API usage has grown roughly 90% over the past year, and token usage has soared 50-fold. That surge has happened even as prices fell — OpenAI cut its GPT-5.6 Luna API pricing by 80% in July, dropping one plan from $20 per million tokens to $6 and another from $1.2 to $1.2.

This dynamic is fueling renewed interest in the Jevons paradox, an 19th-century economic concept that describes how efficiency gains can increase total consumption rather than reduce it. The foundation's executive director, Manick Sundarajan, points to data transmission and storage as precedents: when the cost of transmitting or storing a gigabyte fell, companies and consumers simply used far more data, not less. The same pattern, he suggests, is now playing out with AI tokens, where cheaper access prompts broader and more complex use.

Sundarajan says the real challenge is no longer just about reducing token costs, but about how organizations manage, govern, and apply AI across their operations. As companies weigh open-source models like Meta's Llama and Google's Gemma against low-cost alternatives from Alibaba's Qwen and Moonshot AI's Kimi, the pressure to control AI spend without stifling innovation is likely to grow.