Token prices keep falling 10x a year, but enterprise AI bills still blow past budget
vaibhavbetter · x · 2026-07-21
The post argues that model matching, not cost-cutting, will remain the most effective way to optimize inference spend as usage keeps rising.
It cites the idea that token prices have been falling roughly 10x per year, with a unit of inference that cost $60 per million tokens in 2020 now costing pennies. But despite that drop, many enterprises are still over budget on AI because demand expands faster than unit costs fall — a Jevons-paradox-style dynamic.
Related event: Falling AI Costs Trigger Jevons Paradox, Driving Up Compute Demand(3 posts)→
More from Infra
- 12 KV Cache Reduction Techniques Every AI Engineer Should Understand, Explained — blaizedsouza · 2026-09-11
- The shadow GPU capacity market is formalizing, with Meta selling excess compute to outside buyers — DavidLinthicum · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11