The article analyzes the rapid decline in AI token costs and projects that LLMs will become embedded infrastructure within 1-2 years, with frontier-quality local models on commodity hardware arriving in 3-6 years. It explores supply-side and demand-side Jevons Paradox effects, questioning how investors will recoup costs as tokens become cheaper than tool calls.
Background
AI token pricing has been dropping dramatically as GPU efficiency and model optimization improve. The article draws on hardware efficiency trends and cost trajectories to project future accessibility of frontier AI.
- Source
- Lobsters
- Published
- Sep 23, 2026 at 12:28 AM
- Score
- 7.0 / 10