Re: Why is INTC turning negative in this big Christmas rally?
I think the price went up when it became clear that the P4 was functional, was for sale in the marketplace, and offered higher MHZ than anything this side of a DSP. But then some of the fund guys and retail buyers started digging a little deeper. P4 has a very, very, big die, and the thing to keep in mind, is that as die size goes up, die cost rises faster than die size. Yield is determined by the number of good die that result from each wafer. The current Celeron is 90mm2 and the current P4 is 217mm2. A little over 300 Celerons fit on an 8" wafer, while only about 125 P4s fit on the same wafer. Intel is said to have a very good process, yielding over 70%, perhaps getting close to 80% on existing (Celeron/P3) chips. This means there are typically about 60 to 90 defects on a wafer - each defect results in one lost die. With P4's much larger die, there will be more overlap of defects (where a second defect occurs on a chip already lost due to a first one) so the number of die lost will be in the range of 50 to 70. The result is that Intel gets 240 P3s or Celerons from a wafer (they are the same size) but only 55 to 75 P4s. If the direct cost of a Celeron / P3 is $25, then the direct cost of a P4 is $80 to $100+. Now add in the $60 Rambus subsidy Intel is including for Q1 and Intel's cost per P4 is around $150. AMD's Duron cost is around $30 and its Thunderbird cost around $35 and they are about to come out with a shrink which will lower costs further.
Intel has a strategy of trying to keep its competitor's supply of "oxygen" (capital) low. So they need to make plenty of Celerons and sell them cheap. But they have to make money somewhere. Their quarterly costs are nearly $6 Billion, and the non-X86 parts of Intel must be subsidized since they lose more than a half Billion per quarter, so the Intel Architecture Group must make big profits.
So Intel has to keep margins high on all but the least expensive of its processors.
A looming factor is the big change in PC infrastructure costs that started happening this week. Integrated motherboards for the Duron / Athlon started showing up at retail (I'm sending this from such a system) and this means that there is a huge change the costs seen by the market for Celerons vs. Durons.
Old Distributor Prices
$70 Celeron + $75 Motherboard = $145 cost
$50 Duron + $125 Motherboard = $175 cost
New Distributor Prices
$70 Celeron + $75 Motherboard = $145 cost
$50 Duron + $75 Motherboard = $125 cost
The Duron substantially outperforms celeron, and has been moderately successful at retail even under the old cost structure. Under the new cost structure, celeron will be a lower volume part (even if celeron is given away free - the cost to move to a better performing duron system would only be $50 - a price many are willing to pay), and Intel can't afford to shift too much volume to P3 (which remains the bulk of their line for the rest of the year) or they will impact the corporate desktop market which would bring their ASPs down to the level of AMDs - at which point AMD still makes money, but Intel starts losing $1 to $2 Billion per quarter.
Bottom line is, Intel is going to make less from Celerons and low end P3s this year than they did last year. So that seems to suggest that they should use those otherwise nearly valueless wafers for P4s. Trouble is, costs are high for P4, and the mediocre reviews have keep demand depressed, so if they try to move too much product into the market they'll quickly push prices down to near cost on P4 making it less profitable even than Celeron.
AMD has been threatening to enter the mobile and SMP markets for many months, but so far threats are all that anyone has seen. So the mobile and SMP markets will continue to be profit centers for Intel (unless something changes). But notebook sales have been hit hard recently, so the mobile/SMP markets alone are not going to be enough for Intel in event of a collapse in ASPs for their desktop parts.
Athlon continues to put more and more pressure on P3, so it's hard to see just where Intel can go. In Q2 the Rambus subsidy for P4 goes away, and that will help P4 costs enormously, but it won't help demand much (lower RDRAM prices could help, though). Late Q3 or Q4 Intel's costs will come way down if the triple move to .13, copper, and low-K goes off without a hitch. The other issue is that AMD will be sampling a mainstream 64 bit processor in the Q4 time frame, and if the market finds that as compelling as it did the introduction of a 32 bit mainstream processor back when 386 was introduced, then Intel faces a whole new set of challenges.
Until then I think they're damned if they do and damned if they don't.
Intel has a business model that is based upon $200 ASPs minting money for them and $150 ASPs leaving them as an OK business and they are facing a competitor who's business model is based upon minting money at $100 ASPs and being an OK business at $75 ASPs. Intel has to be very, very, very, careful about letting the perceived value of its mid range and high end processors drop, and it seems to be happening anyway. AMD's $100 ASP business plan looks like an almost inevitable success story while Intel's $200 ASP business plan looks doomed.
Dan |
