Presumably they mean that the CPU resources are over-provisioned, meaning that the virtual CPUs allocated to VMs have to share a smaller pool of physical CPUs. If the VMs have a lot of idle time, this can work well, but if your VM suddenly needs more CPU, the processes on your VM might need to wait for a physical CPU, as physical CPU cycles that would normally be available to you have been “stolen away” by processes running on other VMs.
He called Zuck a cuck?