Unless you have a media intensive application (ex: video, music, etc...), hosting isn't really that bad. I would venture to guess apps like "X Me" and "SuperPoke" would do with a single powerful dedicated server, since theres really not much going on. Also, Facebook hosts the images on any profile, so any bandwidth there is only needed until their proxy server picks up the image.
If your app isn't successful then there's nothing to worry about.
But what if your app is super successful? How would you ever be able to secure enough servers to meet the demand without burning a very deep hole in your pocket?
What we need is a Y-Combinator like incubator for Facebook Apps - a company which will host your app and help you scale it up for a chunk of equity.
- You don't get a static IP address, so if your instance fails and needs to reboot, you might end up with a different IP address.
- If your instance fails, it also loses all the data stored in it.
- There is no hardware load balancing.
With that said, you can probably use it to complement your other servers by doing some of the batch processing.
(a) Economically hosting is a commodity. It's not an area where we can add any special value. We don't do hosting for the same reason we don't supply office or living space: it's better for founders just to buy these on the open market.
(b) Hosting is a technical problem, which it's good for founders to have to solve. We're happy to deal with all the paperwork of incorporation, because you really don't have to know about that stuff to run a company. (I haven't read half our legal documents myself.) But startups do have to understand how to manage servers.
For now, worry about the real issue: "What is a problem that I can relate to and solve for Facebook users?"
The hardware: Two modestly equipped boxes from ServerBeach.com -- and it ran super smooth.