Albert, Re: "it would be no brainer a while back as well but they did not do it, why? They have more than enough capacity....."
Not necessarily. It all depends on demand, and I'm guessing Intel's demand forecasts have changed considerably over the past 6 months.
Let's say you have Y chips that the market demands, and that can be defined by the number of wafer starts times the number of chips per wafer, X. Let's fix this amount, for the sake of argument, at 21,000X. Let's also assume for the sake of argument that Intel can fit twice as many Pentium III die per wafer as Pentium 4. Let's say that 6 months ago, Intel could have filled it with:
A) 10000 Pentium IIII wspw / 1000 Pentium 4 wspw (11,000 total and 21,000X dies)
B) 8000 Pentium III wspw / 5000 Pentium wspw (13,000 total and 21,000X dies)
C) 6000 Pentium III wspw / 9000 Pentium 4 wspw (15,000 total and 21,000X dies)
D) 4000 Pentium III wspw / 13,000 Pentium 4 wspw (17,000 total and 21,000X dies)
Now, let's say that the fabs are currently producing 12,000 wspw, but more money, more energy, some expansion and additional expenses can push it up to 17,000 by the end of the year. For one thing, it's low risk to go with option A, because Pentium III is the current infrastructure, and it will definitely sell over time. In a situation where competition is small, it makes sense to maximize profits, so Intel would start with A, and slowly move their way towards D by the end of the year, at which point there would be an eventual cross-over in volumes sold.
Now, let's say demand is much smaller, such that equals Y - 20%, or 16,800X. Now, Intel is faced with some different choices. By this time, we assume that Intel has ramped up to 15,000 wspw.
A) 7000 Pentium IIII wspw / 2800 Pentium 4 wspw (9,800 total and 16,800X dies)
B) 5000 Pentium III wspw / 6800 Pentium wspw (11,800 total and 16,800X dies)
C) 3000 Pentium III wspw / 10,800 Pentium 4 wspw (13,800 total and 16,800X dies)
D) 1000 Pentium III wspw / 14,800 Pentium 4 wspw (15,800 total and 16,800X dies)
Now, the choice becomes a little different. Intel would like to get the fabs running at near capacity, and they still intend to ramp production. Demand has fallen 20%, so they begin to change their fabrication plans to accelerate Pentium 4 production. Therefore, C becomes the obvious choice, while they eventually move to D once capacity is larger. Eventually, if demand increases, which Intel assumes it will, their .13u plants will have begun manufacturing smaller dies, their 300mm wafer plants will also increase the number of dies per wafer, and their expansion will allow for more wafer starts per week. This is how I think of it. What do you think?
wanna_bmw |
