The only way to determine the amount of time realistically is to perform a trial restore - not a bad thing to do at some point to ensure that the backups are viable. I am sure that you can think of many of the performance affecting factors yourself - ie hard disk performance, I/O channel performance, CPU performance, memory quantity and performance. There are just so many variables that it is impossible to predict without a lot more information.
Also, would you be starting with a bare metal server as in disaster recovery and having to build a basic environment that can handle the decryption?
Don't ignore the human factor - will there be skilled resources available on tap 24/7 or will there be a delay in sourcing manpower? If the event that causes the restore requirement also causes serious or fatal injuries to the entire IT department, and destruction of the entire building, how will that affect recovery times?
Apologies if this sounds facetious but these "how long is a piece of string" questions from box ticking officers really do not show any understanding of what may be involved.
