Its funny, if you invented a coffee maker that grind beans, air-ram them into a filter, and make a fresh ground cup of coffee - (as happened in the past 5 years) would you then start building electric generators and learn about power generation and distribution to handle it?  Or would you design it to be 110v60cycle and be done with it?

Thats where the web is now - a few big players have massive grids (google, AOL, Yahoo, MSN, a few others.  ) And those are tought to run, ask salesforce.com how tough.

So new entrants to web technologies have to know how to build and scale a grid computing device.  Scaling things is hard. Google the airbus 3xxx if you want to learn how hard things are to scale.

So whats coming?  On demand grids, where if you write a 100v60cycle application, you just set it out there and let it run, for 10, or 10k customers.

And my bet is google will beat everyone to it.  Write you great next web 2.0 application.  Don't worry about scaling it for 10, 10k, or 10m users.  Google will host it for you.  They want those eyeballs and can make money showing the ads, whether you run the servers or they do.

Cause servers are getting cheaper every day.  The cost of marketing (superbowl ads a great example) goes up every year.  Plot the curves - cost of CPU cycles and band width, rates for internet ad clicks * internet users.  Its hard to not make an infinite amount of money quickly.

Google knows how to scale massive apps on the net.  And they will start to monetize others apps, not just thier own.  My guess is before 2008.