bobs
t's going to be difficult for AMD to shift too much production to AMD64 products without doing damage to to the other markets it has fought to create
I really don't think there is any difficulty anytime soon. Look at diesize inventories, which are work in process valued at manufacturing-costs of the last batch of these dies. Talking about K7, it should be in the ballpark of 15 Dollars for a fully tested workable die.
AMD reports its WIP inventories quarterly. It would need estimates on the portion of dies in FAB30, Fab25, JV1, JV2 and JV3 plus Flash-die inventories thereof to come to an approximation how much AMDs CPG-die-inventories value is.
A wag of 150M in this respect would give you 10M dies to work with for a year after the last K7 waferout. AMD is known to built huge piles of inventories before phasing out a process. (Remember how long Fab25s Duron was around?)
The crucial point is how many of inventorized dies will be worth to be packaged to a product, respectively whether you can make 30 Dollars or more on the packaged chip or not. This would depend on the stucture of these dies in terms of characteristics (bins e.g). in comparison to the expected sweetspot in future quarters determined by competition. I see AMD able to sell virtually all of these dies in the current competitive environment which is unlikely to change much next year - beyond that it is not at all impossible that AMD would be able to work with dies as well which are currently considered to be not saleable and consequently have an inventory value of zero. This happened in Q4/03 and Q1/04 for a million of CPUs in each quarter.
Bottomline: Marc Lipacis pointing to manufacturing constraints in 2005 without even mentioning inventories is missing something. I don't know whether he does not know about it or he does not want to write about it.
K.
News 
Market Data 
Discover 
AI
