Cloud Computing's Varied Economies of Scale
The value proposition for IT has always been economies of scale. IT enabled the growth of businesses by allowing for massive transaction speeds and volumes. In computing's early years, the expense associated with data centers resulted in time-sharing -- a few businesses investing in mainframes, and others buying processing and storage from them.
Microprocessors and storage in the 1980s resulted in a backlash against centralized data centers; since hardware was so cheap, why not localize IT and ignore any savings from reuse? However, by the 1990s it was clear that despite low-cost processing and storage, IT costs escalated dramatically due to the management and maintenance associated with highly distributed, unshared, and under-leveraged resources.
Cloud computing represents a return to time-sharing, but leveraging the advances of the last 30 years. All of the service models enable some economies of scale. In IaaS, the data center (real estate, power, cooling), the processing and storage hardware, and the firewalls and networks, are all shared among the cloud provider's clients. Fixed costs are distributed, resulting in savings that can be shared among the provider and clients. At the opposite end of the spectrum, SaaS providers run complete software applications designed for multitenancy -- adding a client to Google Apps does not require Google to roll out new servers, databases, and software specifically for that client. The economies of scale for such multitenant SaaS solutions can be enormous, often in excess of ten-fold savings.