The iPhone 4 cost breakdown puts the (16GB) flash memory at only $2.5. And since 64GB USB flash drives retail for $50 and up, I'm inclined to believe that the cost of it for Apple is actually $10 or less.
Only point 2 could happen. 1 can't happen because the processing algorithm will be the same, software is software. Only thing that can influence this is the quality of audio recording, but I doubt 4S has some magical microphone. I'm guessing you're basing point 3 on what that guy said about the GPU coding the voice into a small pack of data. That has to be done other devices as well, one way or the other. The server won't know how to respond to anything else other than what it's programmed to accept. It's not like you can send it an MP3 file and expect it to work with it.

