It doesn't have to be fully ramped to make a huge difference if it is fully 65nm. First, let us assume that the 17% or so of the market is the best that Fab30 can do. Fab36 uses 300mm wafers, each of which can have 2.25 times +/- of a 90nm wafer at the same geometries. In addition, at 65nm, that roughly doubles the number of die per wafer. Assuming that Fab36 is only running 1k wafer starts per week by the end of 2006, that would mean that AMD can roughly double its market share with the two fabs alone.
Me...
That's pretty much how I see things going. You and I have both seen AMD stumble, more than once, over the years when introducing new process technology. However, there is ample evidence that things could be different this time. The "E" stepping certainly was a level of process improvement I don't think AMD has ever experienced. Clearly, the move to jointly develop manufacturing processes with IBM was one of AMD's defining/differentiating moments. The synergies, for both companies, seem to have been enormous.
I'm sure you're hoping, as I am, that the good news continues with fab36. If AMD can reduce the time lag between INTC and itself in doing the move to 65 nm, to less than 6 months, that would be amazing. Again though, I'm still not convinced that INTC's 90nm process is anywhere as good as AMD's so perhaps I'm doing needless hand wringing. Certainly, everthing Hector has said for the last year seems to indicate that fab36 is ontime and meeting process goals.
Perhaps the best empirical indicator of just how well AMD is doing is the lengths, the suit eneumerates, that INTC has had to go to protect market share. It certainly looks like the demand is there, all AMD has to do is supply the goods.
Once AMD can prove it has the ability to supply whomever with all the product they need, that should embolden a sizeable number of OEMs to risk incurring INTC's wrath. It's looking more and more like next year will finally see the payoff we've been so patiently waiting for. All the ducks should finally be lined up and all of INTC's advantages negated.