It’s great to see this conversation start emphasizing the enterprise opportunity for cloud vendors. Up to now so much attention has been given to consumer services, or business apps delivered aaS. But there is a huge opportunity in servicing mid-sized enterprises whose myriad datacenters inefficiently utilize gigahertz and watts and, much more importantly, are ill-equipped to respond to business needs.

On the first point, scaling way up is normally not an issue for these enterprises. That’s not to say that there isn’t a reason for the datacenter to scale way up; on the contrary, this is absolutely necessary to create the scale that can more easily be priced as a variable cost. But this is an internal issue whose benefits manifests only in the unit-costs of the megahertz and megabytes delivered. I don’t need to scale my applications up; we have a fixed number of users that may grow slowly over time but will never require “internet scale”. However, I would love to be able to avail myself of economies-of-scale that I cannot muster myself but are available because of a massively scaled up compute infrastructure (owned by someone else) that can be carved up and sold by the piece for much less than I can do it myself. The Interesting question would be “to what degree can an externally hosted infrastructure wrap up all of the fixed costs and pass them on as a variable, pay-as-you-go pricing that is cheaper per unit than I can provide myself?” (say, with a very well run virtualized infrastructure that I chargeback the business units for).

Secondly, but much more importantly, lowering the unit costs of compute resources may not be the biggest win for pay-as-you-go infrastructure services. Being able to respond promptly to requests from the business units for new value-creating applications may be a much more lucrative reason for someone like me to buy compute resources from the cloud. For example, we have a new supply chain application coming along soon. No doubt a reporting/analysis system will follow shortly after that (unfunded by the original project, of course), and then new functionality will be added by our development staff and THAT will need a few servers, and by then a new financial system will be put on the radar which will need to run in parallel with the original for a year or so on its own set of servers, and then a few integration servers will pop up to shuttle data back and forth, and so on, and so on. We infrastructure managers cringe at requirements like these because we have this big fixed-cost beast from which variable demand must be serviced. Cloud computing allows me to buy the infrastructure when it is needed, provision it effortlessly (“frictionless IT” anyone?), and decommission it when no longer needed

Bring it on!