If they wanted to be really clever, they could also analyze the data when it's uploaded and encode it in a less redundant way... for instance, it's quite likely that large numbers of people will have very similar mp3 files... if you could figure out that 99% of a certain file is exactly identical to 99% of another file, why store that data twice?
But, that gets very complicated... you're just shifting the hard drive costs into developer and cpu costs... so it may not be economical.
But, that gets very complicated... you're just shifting the hard drive costs into developer and cpu costs... so it may not be economical.