Over/under commitment is not really a VM specific thing, although a VM can be affected by it.
We talk about over/under commitment of host resources (CPU, Memory) when deploying VMs to hosts.
A simple example:
A host with 8 CPU cores would be under committed of the VMs it is running only use 4 cores between them. As 4 would notionally be uncommitted, or not used to run VM workload, although in reality the host hypervisor itself also uses these cores to run tasks as well.
When you start allocating more cores to VMs than the host physically has, then this means the host cores are over-committed, and the hypervisor will be required to 'share' the cores as equitably as possible to the VMs, in accordance with any rules and parameters set by the system administrator.
This same scenario also exists for host resources such as CPU clock cycles, and memory. Storage can also be overcommitted, but thin provisioning VM virtual disks and creating more of these on the storage than could be accommodating if those virtual disks were full and grew to their maximum size.
You could also argue that network access could be over provisioned, but this is not really any different to a physical computer network, however it is important to realise that VM network IO consumes host CPU resources, so a VM sending and/or receiving a heavy network load, will be more host CPU resource than just the vCPUs it is allocated, unless limits are set.
