^ Unfortunately all of the above takes $$, and lots of it. As I pointed out in some other threads here, AMD is currently spending something like 35% of its income on R&D, Intel < 15%. Since Intel's income is around 10x that of AMD's, Intel still spends far more on R&D than AMD does. Given the complexity of designing modern-day CPUs, there's generally a direct correlation between the R&D funds spent and how well the ensuing designs work. Granted it's not a perfect correspondence - just look at Prescott-Netburst & Barcelona's initial release for example. However if you look at the overall trend lines you'll see the relationship.
Of course, AMD likes to form consortiums (HTT, SOI process with IBM et al) so you really have to take those R&D dollars into account as well since the results are shared between the partners. And Intel of course spends R&D dollars on SSDs, graphics, process, etc besides CPU designs. However I'd bet a donut that Intel spends more on x86 CPU design than AMD and its partners do.
If AMD could have brought Bulldozer forward to next year, instead of sometime in 2011 which is what AMD is saying, then they would stand a better chance of regaining the top end. But the longer they wait, the more advances Intel makes and with Intel's tick-tock schedule, AMD is hard pressed to keep up let alone cut down the gap.
That's not to say that game-changers are not possible - somebody working in Santa Clara might discover a new super-duper veeblefeltzer transistor tomorrow (well maybe Tuesday since tomorrow & Monday are days off from work
). But CPU & silicon semiconductor tech is mature now, and the chances of such "eureka" discoveries are correspondingly more remote.
Ditto for process tech - Intel is, and is likely to remain, a full year ahead of AMD or GF. AMD has confirmed BD will be the first 32nm CPU for them, so that means 32nm SOI is probably at least a full year or maybe 18 months behind Intel. The die shrink & process maturity yield increases mean that it becomes cheaper to make CPUs.