So if I understand it correctly, this is now based on total allocated RAM to VMs? IE I shouldn't be able to allocate for more than 32Gb of RAM to all the VM's on the server?
That is correct. However the 32GB of vRAM is a soft limit. But, if you oversubscribe vRAM you may not be able to power on some guests.
Just remember, when you accept the EULA, you also agree to go beyond the 32GB of allocated vRAM