Token economics is the cost discipline of AI products: modeling spend per task (input + output tokens × model rates), engineering reductions (routing, caching,
The cost discipline of AI products: modeling spend per task (input plus output tokens times model rates), engineering reductions through routing, caching, smaller models, and shorter outputs, and pricing products so value captured exceeds inference cost with margin.
Because it turned 'AI feature' into a P&L line. Teams track cost-per-conversation and gross margin after inference the way SaaS tracks hosting; investors probe it in diligence and engineers own it in dashboards.