I’m not sure if it’s just my imagination, but the advance in processor speeds seems to be a bit sticky at the moment, with Intel not quite able to offer 4Ghz machines and AMD doing goodness only knows what (not that I’m unimpressed with their processors, but their labelling is a confounding factor). So what’s likely to happen next? More processor speed increases? New architecture running x86 legacy stuff in pure emulation? Mass migration to 64 bit? Something else?
Hmmm… I’m not sure, but I just got a new home desktop with a hyper-threaded 3 gigahertz pentium, and I’m quite impressed with it - though it takes a bit of fiddling to figure out how to get the best results out of a single processor that acts like it’s a dual processor LOL. 
(multitasking here I come!!)
It’s gotten increasingly difficult to increase clock rates and most of the “easy” architectural enhancements have already been made. The next big step is probably going to be pushing chips with multiple cores, such as two processors on a single chip. It’s a relatively straightforward thing to do if you have a big enough transistor budget on the chip. Multiprocessor systems are already common on workstations and servers.
It has a PowerPC core orchestrating the activities of multiple (in the link it’s 9) “synergistic processing elements,” all on the same piece of silicon. The internal data bandwidth is one BHz (bazillion Hertz). Okay; I made up BHz, but the cite gives it as 6.4 GHz. (BTW memory bandwidth has been a big bottleneck these past few years in graphics and signal processing applications).
There’s some talk recently about IBM’s Cell processor scheduled to hit the market next year.