I believe you are conflating kbits/s with kHz. The latter being the unit of the sample rate. The thing is, no human ears hear higher tones than 17kHz and Nyquist and Shannon thought us that double of that as a sample rate is sufficient to reproduce tones equal or lower.
I am using k in this particular message to mean kbps but it I think it does describe the menu of sound quality options that people will hear.
I know people can't hear sine waves above 17kHz but there are questions about transient response and particularly how accurate you would need to represent phase if you want to replicate how well people can spot the direction of sounds in the real world. (Notably no "surround sound" technology of any kind would help a blind man with a gun shoot as accurately as they can in the real world)