The Pricing Crisis in AI-First Products
Token costs are variable. Customers expect predictability. The math is breaking margins.
LW
Liam WuContributor, The Signal
AI-native startups are running into a structural pricing problem their SaaS predecessors never had: their cost of goods sold scales linearly with usage and is denominated in someone else's pricing decisions. A model price cut from a major lab can wipe out a margin assumption overnight, in either direction.
The response from operators has been a quiet shift toward tiered usage caps, fair-use language, and outcome-based pricing models that decouple customer value from token consumption. Companies that locked themselves into unlimited-usage flat fees are now retrofitting metering they should have built day one.