All the question is not if but when the 1 PFlops will be reached.
We need an additional 400 GFlops. Here is my guess. Q4 2013.

If or when GPU crunching comes at WCG we will be able to make a major jump in performance. I stand by my bet: 1PFlops by 31st Dec 2013.

Ok. Sek, Done as you said. You can start the Road to the 700 TeraFlops.

Here is an article about the fastest super computer http://en.wikipedia.org/wiki/K_computer. As one can see it is capable of doing 10,000 Teraflops. I believe that each result is accomplished by at least two computer therefore cutting the effective speed for WCG in half. So I believe that we would have to multiply our results by at least 30 to be an equivalent of the K super computer. Here is another article about a supercomputer http://www.vtnews.vt.edu/articles/2012/01/010512-engineering-hokiespeed.html. This supercomputer is capable of doing 455 Teraflops which I believe is better than what WCG does. Now this supercomputer was bult for only $1,400,000. Now lets say that only 100,000 of the present members decided instead of using their computers to volunteer, they instead decided to donate $20. They would probably save the $20 in less than 6 months of electricity costs. WCG could than buy the Hokie supercomputer and still have $600,000 for operating expenses. I truly believe that the day of distributed computing will end very soon. I expect it wil be withing the next 5 years.

The whole original premise of Boinc was not to replace supercomputers but instead to get people to install a small program to donate their unused computer cycles to something, spreading out the costs. I think that premise is still a valid one as running a supercomputer is not cheap, nor is the team of people needed to keep it humming along. And of course a supercomputer is fast TODAY but tomorrow it is as outdated as our desktop pc's. One advantage projects have is that we users are always buying more 'stuff' and that is free to the project, they just get the benefit of it!