Usage-based pricing, paying based on actual consumption, API calls, compute time, data processed, rather than a flat seat or tier fee, has become increasingly common in SaaS, particularly for infrastructure and AI-related products. It's often presented as a more fair and efficient pricing model, since customers pay in proportion to actual value received rather than a flat fee regardless of usage. This framing is accurate as far as it goes, but it obscures a structural shift in who bears cost uncertainty, and that shift consistently moves in the direction of the customer, not the vendor.
Flat pricing puts forecasting risk on the vendor; usage pricing puts it on the customer
Under flat, seat-based or tier-based pricing, the vendor takes on the risk of forecasting aggregate usage across their customer base and pricing tiers accordingly, if some customers use the product heavily and others lightly, the vendor's pricing model is built to average out across that variation. The customer's own budgeting is simple and predictable, a fixed cost known in advance regardless of how usage happens to vary month to month.
Usage-based pricing inverts this. The vendor's revenue now scales predictably and favorably with actual usage, removing much of their own forecasting risk, since they're compensated proportionally to whatever usage actually occurs, but the customer now bears the burden of forecasting their own usage accurately enough to budget for it, and any unexpected usage spike translates directly and immediately into unexpected cost, a risk that a flat pricing model would have absorbed on the vendor's side instead.
