Single Point of Failure

"When there's one computer serving the planet - even if it's Google's - that's a single point of failure."

I take your point, but you could argue that what made Google great was tying thousands of cheap computers together to appear like one. The best search algorithm in the world is no use if it hasn't got an easily scalable, highly available system to run on, and Google were the first to do this. So it's not just "one computer", and although I concede that no system can be 100% reliable, Google are pretty good. I doubt there are many sysadmins on the planet who would sneer at Google search's availability.

Another plank in their strategy is AppEngine - it gives every wannabe web developer an easily scalable highly available system to play on for free - just as the original PC gave every wannabe computer programmer a system to play on for much less than ever before. I expect most new web applications to be born on AppEngine from now on. Why go to all the hassle of building or renting your own server, let alone a scalable redundant system of servers, when you can use Google's for free?