Bottom‑up AI API revenue models

Updated: 2026.11.20 2H ago 1 sources
Build forecasts from the ground up: estimate usage (Q) by application (coding agents, customer service, research, consumer apps), convert activity into token volumes, and pair with projected API prices (P) to model revenue trajectories. Decompose expected growth into revenue from new applications, deeper use of existing apps, and the offsetting effect of falling per‑token prices. — A rigorous P × Q, application‑level approach materially affects company valuations, investor narratives, labor‑market forecasts, and policy on AI subsidies or regulation.

Sources

Reaction to comments on Anthropic’s AI
Arnold Kling 2026.11.20 100% relevant
Arnold Kling asks explicitly for P × Q token‑based forecasts and a decomposition of Anthropic’s expected revenue growth (new apps vs. usage vs. price declines) and warns that firms will engineer to cap API spending (switch models, cache, compact I/O).
← Back to all ideas