@P.Douglas --
I think that martin.english hit on a few keys points about all-or-nothing vs. right-tool-for-the-job. No one's arguing (or at least, I'm not) that everything should move off-premise. Beyond what martin mentioned, there are other advantages such as elastic capacity during spikes and low- or no-commitment (or CapEx) usage for all sorts of applications. Those are your "augment" scenarios as you mentioned.
Beyond that, there are definite reasons that "cloud" might be better than DIY -- you don't have to focus on ops details like patching as much, you don't have to wait for hardware deployment or manage upgrades, etc -- regardless of the app type.
Again while I completely agree that not everything is appropriate to move off-premise, the redundancy, security, and backup capabilities provided by cloud computing *can* in many cases make it more stable and secure than many on-premise IT infrastructures that don't have huge scale and massively-automated management.
I don't assert that "cloud" will necessarily "save" you tons of money -- it is converting CapEx to OpEx and it's up to your business to strike the right balance financially and security-wise.
Finally, more and more you're seeing a split between "public" and "private" clouds. I personally think it's not as integrated as it needs to be yet, but certainly Microsoft is working in that direction, and companies like Amazon and Canonical have blazed a lot of ground there.
Jason Sherron