PB, thanks for the very detailed response.
* - Probably the biggest factor in the current supply situation is the P858 / Copppermine delay that was disclosed in June. I believe Intel anticipates full recovery from that snowball by the end of this quarter.
That explains the cumine delays/shortages but not the PIII shortages. Right?
* - Remember last year about this time, when all the hype was on the low end?? I don't believe this really had too much to do with the supply situation, but the market conditions of that time had at least some folks questioning how much demand there would be for high end products going forward. Again, I stress, I don't really believe that really has much to do with the current supply situation as bullet #1.
I agree with you because I think there was enough actual experience during the year to give Intel an idea of where demand would be Q4 plus of course the actual orders from their OEMs.
* - As per one of my more infamous posts, by the time P858 was introduced, market demand for the product was significantly higher than what was anticipated earlier in the year. The market picture was considerably changed from the beginning of the year. However, as with previous items, the current supply situation has more to do with bullet 1 than anything else.
You're certain then that demand was much greater than supply for the P858?
* - Regarding fab staffing and shifts: Nothing to do with it. Fabs run 24/7/ Holidays at Intel. The only holiday fab shutdowns I know about recently was in '97, which you noted was a recession year.
And they were running 3 shifts at each fab?
* - Fab space allocation: I don't see this as an issue, per se, except as a matter of overall capacity. Assuming you could do what you assert, i.e., just reallocate fab space at the drop of a dime, something else would have suffered.
MY suggestion was that CPU's bring in the $ so you take space away from flash or some other less profitable division, and give it to the CPU division. Doesn't Intel do space reallocations on the basis of financial reward to the company?
Intel fabs are running full out, with moth-balled fabs being recommissioned, and expansion plans are being accelerated. Fab capacity conditions are determined at points in the past to the tune of years. It is inherent to the industry that there are both gluts and shortages of capacity on a cyclical basis, the reason being is the lead times to facilitate more capacity. Note DRAM capacity cycles for illustration purposes.
Most office buildings/commercial buildings take 18-24 mos to get from groundbreaking to occupancy. Industrial bldgs...approximately 12 mos. I would think fabs would be somewhere in between. Do you know?
I aware of the gluts/shortages. I understand the gluts.....once a building is 40% done, you might as well finish and mothball it. But with shortages, I don't understand the problem; the chip industry has been growing steadily (even with recessions for some time; after all Intel's revs went up even in 1997 albeit more slowly) every year, so these companies should be building even in slow years plus usually there is time to build once demand pick up. After the recession of 1997 and demand picked up in Q1, 1998, Intel should have been building strongly in '98 to accommodate growth in '99 and '00.
* - re: "Why weren't non fab buildings that take less time to build put up to free room for cpu production?"...I don't understand this question. CPU's currently have to be built in $2B facilities. You can't build them in a garage.
I understand about CPU's. But aren't their other Intel products whose building requirements are much less i.e. flash, chipsets and the like? My intent was to free up existing space for CPU's by throwing up their buildings on a fast track basis.
* - Capacity planning isn't a science; it's the Holy Grail. There is some crystal ball work with this stuff. It is extremely difficult to bring online the exact amount of capacity needed, the the exact time it's needed, as the decisions that facilitate capacity infrastructure are made months and years in advance by function of the lead times involved. I.e., Dresden should be online NOW by this line of reasoning. Obviously it's not..
I understand that planning is difficult but it is unusual to have a company with the maturity of an Intel and the ability to execute to be so off on its capacity needs. While I feel they were over done, the comments of GTW were a serious embarrassment to a company of Intel's stature. It seems given the extent of Intel's infrastructure, there would have been a way to avoid such an embarrassment.
BTW I interpret Intel's apparent taking of the high road re GTW, a way of getting the controversy to a quick death rather than the response of a wrongly accused vendor.
* - Summary: The biggest factor in the current supply picture for P858 is the effects that snowballed for the disclosed 2 month delay. Other bullets are presented in an effort to try and illustrate the difficulties of capacity planning and implementation, but are not the biggest factors that would answer your question, IMO.
I am correct when I say this only effected the cumine shortages and not the PIII ones?
ted
|
