Does audio encoding happen on the client?

We'll be using FMS in an app where users will record their audio and then save it to the FMS for later playback. I'm concerned about the load on the server when multiple users will simultaneously be recording audio (we'll have up to 300 simultaneous users). Does the audio encoding happen on the server or on the client machine? If on the server, I'm concerned that the server may not be able to process all of those audio streams. Can anyone provide any information?