>Actually, I was thinking as much of the temporary drop in capacity as lines are converted to .13µ as any online .13µ capacity.<
For the most part, Intel is building new fab capacity for the transition to 0.13-micron, not converting old fabs.
>If so, the (temporary) reduced capacity of Intel (caused by switch to .13µ, as well as increased P4 production) could lead to a supply-constrained situation for Q4.<
The transition to P4 is much slower than Intel anticipated. Intel will be bringing at least one 0.13-micron fab online to produce Tualatin, increasing their wafer starts by about 15% versus Q4 2000, while AMD will be finishing the ramp of Fab30, for about a 25% increase in wafer starts from Q4 2000. Add to this that Intel's new fab will have more die per wafer, and both AMD's and Intel's percentage yields should be increasing, and it is fairly obvious that the market growth from Q4 2000 to Q4 2001 will have to be in excess of 30% to result in a processor shortage.
>Additionally, AMD might get a few first tier OEM wins for the AthlonMP, as well as many more for the mobile Athlons and Durons.<
Neither of these is likely to happen. To begin with, AthlonMP is going nowhere fast. By the end of Q3 even the small form factor notebooks using the small package Athlon 4 will be established, and the shift will be the opposite direction throughout Q4 has mobile Tualatin begins to reclaim territory.
>Hector has claimed Q4 for .13µ, which implies (at minimum) volume availability in Q1. So why would Intel's (comparatively slow) .13µ ramp hurt AMD's ASPs in Q2?<
By Q2 2002 Intel should have overcome all their setbacks in the 0.13-micron ramp. And, while Intel does not seem to be getting as much out of using copper with bulk silicon at 0.13-micron, AMD anticipates that they will get even less for Athlon. Of course, with the shift to SOI in H2, AMD's ASPs could begin to make a recovery. Nevertheless, I anticipate that the lowest speed grade Clawhammer will be introduced at under $200, possibly as low as $149, and even the fastest speed grade desktop Barton will also be priced under $200.
Of course, Intel could be the one that is really hurting by then. Assuming AMD gets about 20% speed grade improvement for Athlon out of SOI, and Athlon gets about 10% less speed grade improvement from 0.13-micron than P4, the peak speed grade of Northwood relative to the peak speed grade of Barton would be 10% less than the peak speed grade of Willamette relative to the peak speed grade of Palomino. With desktop Tualatin being clobbered by even Thoroughbred, the speed grade advantage of P4 decreasing, and the public becoming more aware of the performance issues of P4 as it begins to dominate the market, imagine what Intel's ASPs will look like. |
