We present a model that suggests possible explanations for the observed proliferation of “pay-per-use” (PPU) business models over the last two decades. Delivering “fractions” of a product as a service offers a cost advantage to customers with lower usage but requires extra delivery costs. Previous research focused on information goods (with negligible production costs) and predicted that PPU, when arising as a differentiation to selling in equilibrium, would fundamentally achieve lower profits than selling. We extend the theory by covering goods with any production cost in duopolistic competition. We show that PPU business models can be more profitable than selling (especially at midrange production costs), as long as their delivery costs are not too high, a requirement that is more easily fulfilled as new technologies reduce these costs. Moreover, if firms are imperfectly informed about their customers’ usage profiles, PPU’s effective pricing of customers’ varying usage offers an additional advantage over selling. This requires companies to employ accounting methods that do not inappropriately allocate production costs over stochastic usage levels. If PPU service provision suffers from queueing inefficiencies, this does not fundamentally change the relative profitability of the PPU and selling models, provided that PPU providers can attract sufficiently high demand to benefit from pooling economies. This paper was accepted by Charles Corbett, operations management.