Ke-Wei Huang1 and Arun Sundararajan2
Working Paper CeDER-05-26, Center for Digital Economy Research Leonard N. Stern School of Business, New York University November 2005
Abstract: On-demand computing provides a new way for companies to manage and use their IT infrastructure. This model of corporate computing radically changes the way companies pay fortheir IT infrastructure, basing it on "pay per use" rather than on the ﬁxed infrastructure investments such companies are accustomed to. A clear theoretical understanding of pricing on-demand computing is thus central to the viability and growth of this nascent industry. We contribute towards such an understanding in this paper by modeling the optimal pricing of on-demand computing while takingfour critical factors into account: the costs of deploying IT in-house, the business value of this IT, the scale of the provider’s on-demand computing infrastructure, and the variable costs of providing on-demand computing. Three distinct pricing models emerge as optimal among all possible pricing functions for on-demand computing. These models describe when volume discounting, free usage and demandcaps should be used to manage demand appropriately and proﬁtably. We also outline a likely path that the transformation towards on-demand computing will follow — under which lowusage customers are targeted initially, followed by a broadening of the market, and ﬁnally, a focus on proﬁting from inducing adoption by high-usage customers — and prescribe how the associated pricing models should evolveappropriately3 .
JEL Codes: D42, D82, L12, L86.
44 West 4th Street, KMC 8-185, New York, NY 10012. firstname.lastname@example.org 44 West 4th Street, KMC 8-93, New York, NY 10012. email@example.com 3 We thank seminar participants at New York University for their feedback. The usual disclaimers apply.
The emergence of on-demand computing promises to transform theway corporations buy and manage their IT infrastructure. This model of computing, also referred to as "utility computing", or in the speciﬁc context of software, as "apps on tap", may shift IT infrastructure from being a fragmented capital asset to being a centralized utility service. The viability of on-demand computing has been facilitated by two related technological developments — grid computingand Web services. Together with the widespread availability of Internet bandwidth, they make it technologically feasible for corporate buyers to "rent" key parts of their IT infrastructure — servers, data storage, isolated software applications, and integrated software solutions, for, among other things, salesforce management, CRM and retail fulﬁlment — from large-scale utility providers, ratherthan deploying and running these parts of their infrastructure in-house. Current spending levels for on-demand computing are still a small fraction of IT corporate budgets; however, this fraction is widely projected as growing rapidly over the next decade. During this nascent stage in its evolution, it is essential for providers of on-demand computing develop and implement pricing and migrationstrategies that make the transition appropriately gradual and reliable. An inappropriate choice of pricing that is based on usage could either lead to excessive inertia in migration, or alternatively, to excess demand that providers cannot fulﬁl proﬁtably or scale to meet reliably. Either scenario could easily kill early innovators in on-demand computing. The importance of a careful and judiciouschoice of pricing models, critical during this transition in corporate computing, motivates our paper’s objective: to develop robust prescriptions for pricing on-demand computing. We identify and model four aspects of corporate IT infrastructure that aﬀect it’s pricing: (1) The cost of buying, deploying and maintaining the infrastructure in-house: Since this 1
option is always a viable...