The Upcoming CPU Glut

I’ve mentioned this before, but I thought it a good idea for you to hear it from Jerry Sanders’ mouth.

All comments from
AMD’s 2001 Third Quarter Conference Call. The complete transcript can be found at
its manual

“. . . [W]e have a magnificent factory in fab 30 which is ramping to full production and we are converting to 130 nm technology. We think this is our secret weapon if you will. It can produce over 50 million units a year in the 130 nm technology and I will remind you that we have an extremely small die on the Athlon product, only 80 sq mm which is dramatically different than our competitors die, so our issue there is ramp that production as quickly as we can and move to 130 nm, highly desirable products. . . . We’ve got an extraordinary asset in fab 30 which will enable us to successfully prosecute the PC processor market in even the most competitive environment”

“It turns out the die size of the Athlon at 130 nm is only 80 sq mm. I just point out that gives us some 315 die candidates per wafer. Intel’s Northwood will be no more than 180 die candidates per wafer, so we think we have a real edge on them on cost structure so they can continue to increase revenue and reduce their operating profits, but I think that is a fool’s errand.”

“. . . [A]s you go to 130 nm, we will be 100% converted to 130 nm by the end of next year. We can produce more than 50 million units a year in that fab. I point out that would be more than the 30% market share that everyone says we will never get to except us. . . . Meanwhile, we do have plans so that when we have to have more than 50 million units per year, we will be on target and I think we will have some more comments on that also on November 8th.”

Now maybe I’m missing something, but I think this guy wants to make 50 million cheap processors a year as soon as he can, don’t you?

AMD is now making about 30 million processors a year. The worldwide x86 CPU market is about 150 million, and will probably pretty much stay that way until sometime in 2003, maybe 2004.

Intel isn’t standing still, either. They’re starting to mumble about introducing .09 micron architecture sometime in 2003.

According to c’t, MicroDesign Research estimates that it costs AMD about $60 to make a current Athlon, and $100 for Intel to make a PIV.

Going to .13 micron is likely to bring AMD’s cost down to around the $35-40 range. Intel will go to .13 micron, too, which should probably bring manufacturing costs down to $60-70. Maybe even lower with these 300mm wafer fabs.

Even if Intel ends up not expanding capacity at all in 2002, AMD will be adding another 13% to world capacity. This would be very roughly the kind of capacity increase you’d expect from Intel in a normal year, not from the 22% marketshare holder. This is not much less than the total increase in world demand for a normal year.

2002 will not be a normal year.

Not only will the world CPU market not grow 13% or more in 2002, it might contract a little more. This means AMD sales are going to have to come out of Intel’s hide. Not just a bit less market share, but literally fewer processors made by Intel.

I don’t think Santa Clara’s going to like that.

So far, the only thing that has sold AMD processors is price. I can hardly see AMD managing to sell 70% more processors AND get more money from each. Only the Yankees can pull off miracles like that. 🙂

Unless the worldwide CPU industry stages a dramatic recovery during 2002 and into early 2003, something’s got to break. It could be AMD’s expansion plans, it could be Intel’s marketshare, but given the current environment and temperament of the participants, I bet it’s going to be price.

A year or a bit more from now, we may be paying as little as $50 (and almost certainly less than we are now) for AMD 2GHz++ processors. This situation will go away with economic recovery, but it could get really rough for both companies for a while.

Email Ed

Be the first to comment

Leave a Reply