Joe, Re: "My theory is that AMD bought equipment as late as they could get away with, and the last of the batches of the equipment were qualified for .13u production"
That's fine, but here is something else that Hector said during the AMD CC.
<Dan Scovel> Thanks. What is you capacity utilization right now out of Fab 30?
<Hector> I would have to say that we are ramping and using it as per our ability to sell the product and we are not selling everything we make and obviously as we convert the factories to 130nm, we are doubling practically the capacity of that factory. So when you look at our wafer starts, which some people do, we are running at capacity. When you look at its die per wafer capability, it is ramping up to be full by the end of the year.
So let's break this down.
- First, he says that AMD is not selling everything they make. Either they are stockpiling (doubtful), or they are demand limited (probable).
- Second, he says that 130nm manufacturing will double capacity. With 4-4.5M units out of Dresden right now, that will allow roughly 8-9M units for AMD per quarter, after they ramp down Austin. Of course, Hammer's larger die (104mm^2 vs 80mm^2) and SOI yields may shrink that number considerably. No wonder AMD is going with foundries.
- Third, Hector says that they are running at wafer start capacity. That means 5,000 WSPW, which is what Dresden's capacity is rating at. Either this has changed, or AMD isn't getting the world class yields that they also claim. Elmer has done the math - he's not just spreading FUD with his poor yields theory. I'm betting that it partially makes sense.
- Fourth, he is saying that die per wafer will ramp up towards the end of the year (2002). This will be mostly due to 130nm manufacturing, and as you point out, AMD is already getting more die per wafer on their early 130nm process than they are getting on their mature 180nm process (quite logical). Moving forward, they obviously intend to double the number of good die per wafer. Using calculations that this thread has done in the past, AMD can easily accomplish that with the smaller die size, and perhaps slightly better yields.
Therefore, given what Hector has said, it is my opinion that AMD does have inventory, partially due to being demand constrained (meaning that Intel parts are still more popular in the market). They also are going to be short of what Dresden is capable of due to lower yields, given that they are at capacity in terms of wafer starts, and they are also short in terms of unit volumes. Suffice it to say, I believe that AMD can continue to remain at their current levels of record volumes, even after the transition to one fab. However, they will not be able to significantly grow past it this year, and they may have problems with capacity next year, but with the foundry work, they may be able to maintain their current levels. This is why it's important for AMD to shift their mix to higher margin mobile and server parts. I don't believe that AMD will try to collapse ASPs (and neither will Intel, for that matter), since they will need the margins to make a profit this year. If the pricewar does continue, though, I see it being tough on both companies. If you have a different interpretation, I'll be happy to hear it.
wbmw |
