More sites are certainly willing to run virtual Master servers, in a single site scenario, when they are willing to accept the chicken and egg situation of... in the event of severe loss that they need to manually (scripted?) rebuild their Master Server (and possibly all / many Media Servers too) and then recover the NetBackup Catalog (via a recovered Media Server) before they can restore any VMs.
The sticky point is usually fibre-channel SAN devices, usually physical tape libraries or VTL tape libraries (i.e. physical disk arrays pretending to be physical tape), and SAN attached disk storage for VMware or Hyper-V. Usually it is quite hard to present physical FC SAN WWPN (for tape or disk) to virtual NetBackup Media Servers. In the past the most performant backups of VMs was to read the SAN storage directly along-side the ESXi hosts, and so it was best if the NetBackup Media Servers were physical servers with real FC SAN HBA.
If the DR protection model for the backup data is replication, and everything is disk (or Appliance) based, then there should be even less need for physical Master and Media Servers, as you have nothing to re-create in the event of DR because it is all already off-site.
These days you can do it all without physical FC SAN, but you have to know your limits and bottlenecks really really well.