When dealing with a Cloud Architecture the ability to scale out becomes available by adding computational resources to process job queues. The application’s work load is divided across a number of servers who each process a job or set of jobs retrieved from a queue.
What you lose in the power earned from Classic Architecture you gain in power though parallelism and cost saving through flexibility in Cloud Architecture. In Cloud Architecture, when more resources are needed more servers are added and when demand has been met and those resources are no longer needed they are removed.
This type of resource management is powerfully meaningful from a cost perspective when the process of scaling out and back is automated.