Business case wise it doesn't make sense to move just to go cloud. When we moved into the rented cloud we did it instead of replacing our old servers that were end of life. This meant we saved money over all because buying and running your own hardware is expensive.
Similarly it wouldn't make sense to move our SQL cluster into Azure just for the heck of it. However when our SQL server license is up for renewal, the move to the cloud will be the best financial option.
In development the story is a tad different. Our operations wizards can't do everything that our developers (both our own but also on open scourge components) require. We're almost 100% Microsoft in operations, but we run 370 different systems and some of these run in Linux, php, nodes, python and so on. By utilizing the cloud, we won't need a certified red hat guy on premise because Microsoft (or Amazon) will handle things like patch management and security updates. Or in other words, it's more expensive to run our applications on a SAAS platform than on our own servers, but it gives us much more freedom to chose and build the applications to support our primary business of meeting the needs of our employees and citizens.
But isn't this just because the initial up front investment for your existing on prem infrastructure has already been spent, and that they are unlikely to all reach eol at the same time? Over time, won't the cost savings of the cloud model make up for the on prem write-off plus the costs of re architecture-ing for cloud? What sort of workloads do you never see moving to IaaS, where the costs of rearchitecturing are just too high? Any opinion on the new VMW on AWS offering, it's pricing, etc?
After that, you opportunistically move stuff as it makes sense. Then squeeze every provider of software and services, and pick off the stragglers by moving to cloud/SaaS.