In an aspect of the disclosure, a method, a computer-readable medium, and an apparatus are provided. The apparatus makes a first determination whether a load metric of a first hypervisor is in a first predetermined relationship with a first threshold. The first hypervisor is one of a plurality of hypervisors of a VDI system. The apparatus also makes a second determination whether an available capacity of a second hypervisor is able to run virtual desktops of the first hypervisor when the load metric of the first hypervisor is determined to be in the predetermined relationship with the first threshold. The first hypervisor migrates virtual desktops of the first hypervisor to the second hypervisor when the available capacity of the second hypervisor of the plurality of hypervisors is determined to be able to run virtual desktops of the first hypervisor. The apparatus also powers off the first hypervisor.