I was shown a design with SQL2012 High Availability where there were 3 nodes in an HA cluster.

The 3rd node was in the cloud and was minimal spec as a DR option. In the event of a full failover the elasticity of the cloud would come to the fore and the node would be spun up with whatever juice it needed.

Low cost node during normal working, high cost just when it was needed.

With SQL2012 the different nodes in the cluster don't have to be the same spec so one node can be specced for OLTP, another for reporting and a 3rd for DR.