With EC2, the end user sends his application to an Amazon data center and orders it to run. With Salesforce, a shared application is already in the data center, and the end user orders an individualized version of it to run. Both of these are strong cloud computing examples and both are much lower cost than their predecessor forms of computing, the packaged applications that Oracle now offers by the hundredweight.
In fact, in a future discussion, I want to describe the multitenant application as one of the engineering accomplishments of our age -- a cornerstone of cloud computing. Its efficiencies surpass those of its predecessor forms of applications and it is the most likely candidate to replace that packaged application in your data center somewhere down the road. Ellison's about face at Oracle OpenWorld was a rare Silicon Valley event where a vehement critic had to eat his words and accept a trend he had been trashing. Too many members of his audience had come to accept the cloud for Ellison to continue in that vein.
But his halting "definition" had less to do with how to deliver the most efficient computing power to end users than with where he thinks the next round of competition is coming from. Looking for an opening, he attacked the cloud's multitenant application -- or at least Salesforce's version of it -- as unsound and unsafe. Ellison as a friend of the cloud is almost more dangerous than he was as a critic. The marketplace is changing, orange is the new pink, and the leopard has changed his tune -- but he hasn't changed his spots.