"now they are going to be at the mercy of one to actually make a CPU."
Sigh, you've been reading Yousef's posts too often. Assuming that AMD does what they claim they will be doing, they will use UMC's equipment to develop their process at the 65nm node, not give up process development and go strictly over to UMC's as Yousef likes to pretend. Yeah, AMD will be using UMC at 130 and 90nm for fabbing chips, but they will also be using Dresden at those nodes, I would guess for the higher value chips. At 130 and 90nm, UMC is there if they see the demand expanding to the point that Dresden cannot handle it, and if that's the case, then even if they only make $1 per chip on the UMC output, it's worth it.
Think about it. If AMD really does produce a ClawHammer on 90nm and ship it by mid-2003, they have 400 potential chips per 200mm wafer. Now assuming they only get around 50% good and saleable die per wafer, they should get at least 200 good die per wafer at 90nm. Assuming 4k wafer starts per week, that is around 10 million ClawHammers per quarter from Dresden alone. More, of course, if they yield and bin higher and/or have more wafer starts. If they use any capacity from UMC, then they will have more. Unless the market really picks up, then to use UMC would mean they have close to 30%, if not more, of the market... |
