Raw Content

How AI Made Pricing Hard Again

The Core Problem

Traditional SaaS companies enjoyed 70-85% gross margins because incremental usage cost them nothing. AI companies face a fundamental economics crisis: they pay LLM providers per use, making growth expensive rather than profitable.

MoviePass exemplified this disaster—offering $9.95 monthly movie passes while paying $10 per ticket. Replit experienced similar pain, watching gross margins collapse from 36% to negative 14% as usage increased.

Five Pricing Models for the AI Era

1. Usage-Based Pricing Charge per token processed (OpenAI’s model). Pure pass-through costs to customers.

2. Seat-Based Subscriptions Still viable when products facilitate collaboration. Companies raise per-seat prices to account for AI costs, though AI productivity gains reduce the number of seats needed.

3. Subscriptions with Overages Include base usage allocations, then charge extra beyond limits. Cursor uses this approach for code editing, where reliability is paramount—users need uninterrupted access.

4-5. [Remaining models paywalled]

Critical Insight

The article emphasizes that pricing decisions now require company-wide involvement beyond finance. Engineering teams must understand margin implications, as AI’s variable costs represent both liability and competitive advantage. Founders building AI products need frameworks for sustainable monetization when their best customers are simultaneously their most expensive.