Earlier quoted context omitted.
It's not unreasonable for drop.io to burn $500k/mo? I run a ten-man startup and I can't fathom burning that much.
Do you host an unlimited number of 100MB files, indefinitely, for an unlimited number of non-paying users?
Some tuning on that, and I'm fairly certain you could run things for less than $500,000k a month -- most files are probably uploaded, downed a few times in a week or two by a select number of people, and then forgotten.
Then again, and this is key here, I have not tried this. It's also something that could be difficult to patch into codebase if it wasn't considered early -- we're all familiar with the problem of coding ourselves into corners, especially when time is on the table.