Pricing trends are recurring patterns in how AI companies price, derived from UsagePricing's 55-company Blueprint corpus. Each trend below is backed by dated evidence and re-tested as the corpus grows.
Updated
AI vendors are increasingly denominating usage in an internal 'credit' currency that floats across features — simplifying the invoice and decoupling the headline price from raw compute, at the cost of unit-economics transparency for the buyer.
A ~$200/month 'Max / Ultra / Pro' prosumer tier has emerged across consumer AI — roughly 10× the $20 'Plus' tier set in 2023 — as vendors monetise power users with priority or uncapped access rather than raising the mass-market price.
Inference vendors have converged on batch processing (~50% off) and cached-input discounts as the standard levers to price-discriminate by latency tolerance — a de-facto pricing playbook for token APIs that smaller app vendors do not offer.
Infrastructure and platform vendors keep adding metered dimensions as they ship features — drifting from one billable unit toward many — while consumer apps hold to a single unit. The metering surface is fragmenting at the infra layer and consolidating at the app layer.
A wave of vendors abandoned flat subscriptions for pay-as-you-go credit metering — but several then re-introduced tier structure on top, converging on hybrid rather than pure usage. The endpoint of the migration is seat-plus-usage, not pure PAYG.
As agents call tools — web search, page visits, code execution, deep research — vendors are beginning to price those actions individually, per call, instead of folding everything into tokens. Early but spreading: the billable unit is shifting from the token to the action.
Per-token API prices fall relentlessly with each model generation and via caching/batch discounts, even as consumer subscription ceilings climb — compute deflates while the price of human-facing access inflates.
Vertical enterprise-knowledge AI (legal, enterprise search, marketing) keeps pricing gated and sales-led, bucking the corpus norm where 40 of 43 companies post public prices. The more a product sells into regulated or seat-heavy enterprise buying, the less likely it is to show a number.
Charging per resolved outcome — not per seat, message or token — is the most-discussed and least-adopted pricing model in AI. The corpus has one clear live adopter (Intercom Fin, per resolution); it's a signal to watch as agents make outcomes measurable, not yet a pattern.
No trends in this view.