The cloud will eventually mature.
At present I think of it as Pay-As-You-Go computing. If you have a facility that you need regularly or occassionaly but not all the time then it can save you a lot of money.
If you are using something 24/7 then it can work out very expensive.
One challenge with data is that it is shared. You may start off with a pool of data being used for a specific purpose and a PAYG model is fine but along comes another requirement that radically alters the usage profile and you get a nasty shock when the bills come in!
With data analytics you have to be careful where you host your components. If your analytics tool is co-located with your data then fantastic. If you have a locally hosted analytics tool looking at a cloud source then you are going to be (trying to be) shifting massive amounts of data from the cloud to your local compute resource.
Similarly, if you have a nice analytics stack the identifies customer segments along comes a marketeer that says "great, grab me all the data from segment 'X' so I can contact them! Again, you have got a sudden big data shift operation going on and the bills stack up!
LinkedIn ProfileNewbie on www.simple-talk.com