There's a significant difference there. Beer costs money to replicate, programs do not. The marginal cost of adding a new user to the Free-Beer program is the price of a case of beer. The marginal cost of adding a new user to your startup's community is a minor bandwidth charge. I'm not completely sure but I'm willing to bet that if you make a good enough service, people will come. With such a small marginal cost per user, you can afford to grow a userbase before you start incurring serious charge, and once you've reached a point where bandwidth costs are actually meaningful, you can almost definitely find a way to make money off of it.
As another user said, the companies listed so far failed for auxiliary reasons. One of them to be wary of is overexpanding, which is basically what Webvan did. The trick is to not do a massive land-grab unless you absolutely have to; Read this- http://www.joelonsoftware.com/articles/fog0000000056.html
If your cost per user is less than the incremental revenue each user brings you, it doesn't matter how little those users cost: You're still following the "we'll make it up on volume" battle cry in the wrong direction.
Bandwidth and server costs are still costs. Whether you split those costs among four customers or four million, you still need to recoup them and show a profit to create a sustainable business.
If you don't have any secondary costs this is true.
A lot of web companies still need to "pay the rent" or other variable costs. Even without that, don't forget that revenue gives you a huge leg up when searching for funding.
As another user said, the companies listed so far failed for auxiliary reasons. One of them to be wary of is overexpanding, which is basically what Webvan did. The trick is to not do a massive land-grab unless you absolutely have to; Read this- http://www.joelonsoftware.com/articles/fog0000000056.html