Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: kash johal who wrote (38278)	5/4/2001 4:11:04 AM
From: dale_laroy	Read Replies (1) \| Respond to of 275872

>What dreamworld are u living in.< Well actually, I was attempting to demonstrate that, even under the most optimistic assumptions, AMD will be in a world of hurting with regards to processor capacity if they do not start ramping a new fab earlier than they are claiming for Fab35. >fab 25 is kaput and on last legs barely producing 1Ghz processors.< I doubt it. Jerry Sanders claims that they are producing no Athlons below 1.0 GHz, and admits that Fab25 is producing Athlons. Even the most optimistic projections for copper interconnects was a 20% increase in speed grade. It appears that a 1.467 GHz TBird could be shipped from Fab30. Even assuming a 20% speed grade boost for copper interconnects this would indicate that AMD could ship at least 1.2 GHz TBirds from Fab25. AMD claims they will be able to ship a 1733 MHz Palomino from Fab30. Applying this 20% speed grade boost for copper figure to Palomino would still yield at least 1400 MHz for Palomino from Fab25. Additionally, by raising the voltage, there is no reason that AMD could not ship the same peak speed grade Morgan from Fab25 as Palomino. Thus, it would appear that AMD could ship at least a 1.4 GHz Morgan from Fab25. AMD has been increasing the peak speed grade of Duron by about 50 MHz every two months, and can probably get away with this until Intel introduces a value segment P4 processor. With 900 MHz shipping now, it will take AMD another 20 months (nearly 7 quarters) to reach 1.4 GHz for Duron. This would be about the end of Q4 2002. And in fact, AMD can continue to ship Duron at a peak speed grade of 1.4 GHz right up to the point where Intel introduces a value segment P4, because the only alternatives will be Celeron (possibly Tualatin) and the Via C3/C4. >Fab 25 has NO 0.13. AMD has no Al 0.13 process.< I have read nothing about a 0.13-micron process at Fab25. But that does not mean that there is not any. I just can not see how Jerry can state that the two existing fabs can meet processor demand without Fab25 going to 0.13-micron. >AMD has already stated Fab 25 goes to flash.< He never said when. >Forget Fab 25 for CPU's by mid 2002.< As I have stated, CPUs from Fab25 should be marketable through the end of 2002, even without 0.13-micron process technology. This would even be the case if AMD could not ship a peak speed grade higher than 1.2 GHz. >Fab 30 is roughly at 50% capacity and its producing about 50% of AMD's units now.< I doubt it. Fab30 is only producing Athlon. There were fewer Athlons produced than Durons. Fab25 is not only producing both Athlon and Duron, but also K6. At best, Fab30 produced one third of the units last quarter. >By Q1 next yr Fab 30 is full capacity.< Agreed. Provided AMD gets their 0.13-micron equipment on time. The final 10% ramp will be at 0.13-micron, with the first shipments from the end of this ramp no earlier than 2.5 months into Q1 of 2002. >So by Q1 they will be able to ship same number of units as today (plus maybe 10% due to yield improvement).< If you are talking about matching today's total capacity level from Fab30 alone in Q1 2002, forget it. At best, from Fab30 alone in Q1 2002 they will be able to ship 70% of the combined number of units from both fabs today. >AMD clearly runs out of capacity by mid 2002.< Hardly, Fab25 will continue to churn out Duron processor right up to the point when Intel introduces the value segment P4. And even after Fab30 is completely converted to 0.13-micron, it will be months before increases in percentage yield manages to fall behind increase in average die size. >Yes 0.13 lowers die sizes. But folks add features making die sizes bigger- its inevitable.< Why not just state that average die size will increase as AMD moves from the 0.13-micron Athlon/Duron product mix to the Hammer product mix? >So by Mid 2002 AMD MUST HAVE outside foundry assuming they are going to increase market share.< No, they must have either an outside foundry or be ramping a new fab by H1 2003. >The only way they do that is by external fab arrangement.< Even an external fab arrangement would not do this. By agreement with Intel, AMD is limited to having 20% of their units foundry produced. The loss of Fab25, regardless of when it occurs, will reduce AMD's unit capacity by at least 35%. Thus, even by outsourcing to a foundry as Fab25 is converted to flash, AMD will suffer at least a 20% drop in unit volume by the end of the conversion versus the point at which conversion is started. >My best guess is IBM/UMC foundry partners in pure buy/sell arrangent followed by some kind of JV fab for fab 35. So from mid 2002 to 2004 they utilize say up to 50% of a UMC 12" fab.< As I pointed out, this can not happen because AMD is limited to outsourcing 20% of their production. At best they could utilize 10% of a UMC 12" fab before they hit the ceiling of the agreement with Intel. AMD has two options for preventing Intel from regaining market share in 2003: 1) AMD could outright license IBM and/or Motorola to produce and sell IBM and/or Motorola branded Duron and Athlon processors, with a possible cap on production, such as IBM being limited to producing one Duron processor for every value segment Hammer processor produced by AMD and one Athlon processor for every Clawhammer processor produced by AMD. 2) AMD can actually buy into a foundry, such as UMC's Fab12 in Taiwan or the Trecenti joint venture with Hitachi, thus escaping the foundry cap. For example, AMD could buy a sufficient share of Fab12 in order to ramp to 2000 300mm wafer starts per week as capacity at Fab25 is converted to flash. In conjunction with outsourcing with UMC for the production of enough Appaloosa processors to match 25% of Fab30's capacity, this could enable AMD to outsource an additional 500 300mm wafers of production to UMC foundry services. >And in 2004 the new JV fab would come on line. AMD has pretty much said they will probably partner for next fab.< If AMD were to use option 1) above, licensing IBM and/or Motorola to match Hammer processor sales with Duron/Athlon processor sales, the non-Intel market share would be high enough at the launch at Fab35 that AMD could utilize an entire 300mm wafer fab simply by changing over to K9 production.