Inference economics / token deflation
Collapsing cost-per-token reshaping unit economics
The cost per token for a given capability fell by roughly an order of magnitude per year through 2024-2025, driven by better models, quantization, and inference chips, shifting the industry's economics from training to inference at scale. This deflation is what enables agentic workloads but also compresses per-token pricing and margins. It is the core financial dynamic determining who profits from AI.
How it fits the stack
Inference economics / token deflation with what it depends on (above) and what it feeds (below). The figure renders as a crawlable diagram and upgrades to an interactive 3D graph as it scrolls into view.
Inference economics / token deflation in the AI stack. Inference economics / token deflation with its immediate upstream dependencies (top) and downstream dependents (bottom) in the AI value chain. Hover a node in 3D, or read the full relationships below.
Graph data (text) — 5 entities, 4 relationships
- Inference economics / token deflation —used by→ vLLM
- Agentic AI / autonomous workflows —depends on→ Inference economics / token deflation
- Cursor (Anysphere) —depends on→ Inference economics / token deflation
- Epoch AI —uses→ Inference economics / token deflation
Depends on ↑ · 1