It appears that the Azure API does not expose functionality to start up an arbitrary number of VMs simultaneously, like OpenStack and AWS do. The result of this is that if you want a fleet of n VMs, you must wait for the Deployer to create n VMs serially* before Ansible begins the provisioning process.
It would be nice if the Deployer could run batchSize threads to create the VMs at the same time to speed up the fleet creation process.