Marginal-cost pricing sets the price of an additional unit at the cost of producing that unit. It is a pricing floor concept: selling at marginal cost covers variable cost but contributes nothing to fixed costs or profit.
Marginal-cost pricing sets the price of one more unit at what that unit costs to produce. It is primarily an economics concept and a pricing floor: at marginal cost you break even on variable cost but contribute nothing toward fixed costs or profit. Understanding it matters because, for AI products, marginal cost per unit is real and non-trivial.
How Marginal-cost pricing works
Marginal cost is the cost of producing one additional unit, the extra compute for one more API call, the tokens for one more generation. Marginal-cost pricing sets price equal to that figure. In perfectly competitive markets, economic theory pushes prices toward marginal cost; in practice, businesses price above it to cover fixed costs and earn profit.
The concept is most useful as a floor and a decision rule: you should generally not sell below marginal cost (each sale loses money), and the gap between price and marginal cost is your contribution margin per unit.
Marginal-cost pricing examples
For pure software with near-zero marginal cost (one more download costs almost nothing), prices far above marginal cost are normal and sustainable. For AI, marginal cost is significant: one more inference consumes real GPU and token cost, so pricing must clear that floor to avoid losing money on every call.
A practical example: if an AI feature costs $0.30 in compute per use, marginal-cost pricing says $0.30 is the floor; the actual price must exceed it to fund the business.
Benefits & when to use it
Marginal-cost pricing is less a strategy to adopt than a discipline to respect. It defines the floor below which a sale destroys value, and the contribution margin (price minus marginal cost) that funds everything else. For AI products, where marginal cost is real, it is the guard against the trap of flat pricing that loses money on heavy users.
It is rarely the actual price (that would ignore fixed costs and profit), but it is the essential input to any usage or value-based price. Knowing marginal cost per unit is the prerequisite for pricing AI profitably.
FAQ
What is marginal-cost pricing?
Setting the price of an additional unit at the cost of producing that unit. It is a floor concept: at marginal cost you cover variable cost but contribute nothing to fixed costs or profit.
Why does marginal cost matter for AI products?
Because AI has real marginal cost per unit, each inference consumes compute and tokens. Unlike traditional software with near-zero marginal cost, AI products must price above marginal cost or lose money on every use, which makes knowing it essential.
Should you price at marginal cost?
Usually no. Marginal cost is the floor, not the target; pricing there ignores fixed costs and profit. The gap between price and marginal cost is contribution margin, which must be positive and large enough to fund the business.
How Credyt handles Marginal-cost pricing
Pricing above marginal cost requires knowing it per customer and per event, which is exactly what Credyt measures. By ingesting vendor costs and attributing them to each usage event in real time, Credyt exposes per-event and per-customer margin, so teams can ensure prices clear the marginal-cost floor instead of discovering negative margin at month end. Explore Credyt →