I've been reading a lot about Google and their technology over the past few years. One thing is becoming increasingly clear: This whole "Google" thing is about much more than search, and I think they are going to become increasingly relevant in the server-based computing space.
Many visionaries have predicted that computing in the future will be "grid"-based, where we just connect our endpoint into some ubiquitous computing environment. (I mentioned this briefly this last July.) In a sense this would be the ultimate server-based computing environment. Not only would we be able to access our stuff from anywhere and any device, but we would have virtually limitless storage and computing power available to us on the back end.
Google's server environment is made up of more than 100,000 individual servers. People have speculated that it includes a distributed fault tolerant filesystem exceeded one petabyte (1000 terabyte) in size. It also most likely has distributed RPC code, network shared memory and process migration. As Jason Kotke pointed out, this is basically a giant supercomputer that doesn't crash.
He also pointed out that this might mean that Google could create an environment where users are able to run actual applications from this grid. Users could have an office application that could be used from anywhere with their own private storage in Google's grid that was always available. Google could facilitate an environment where you wouldn't need your own computer to be productive. They could take the key operating system away from the client device, essentially meaning that all anyone would need would be a thin client.
Of course all this isn't really relevant right now, but I bet it won't take too long (five years?) to start seeing some of these basic ideas implemented. This is yet another reason I'm excited to be in this field...