Brian, thanks for the update. DRAM is a slippery part of the business to assess, I think. First off, there's all the "missing inventory" as discussed in a recent Contrarian column here on SI. If there's really that many pieces (100-150 million) of DRAM hidden in warehouses somewhere, then the current price supports are artificial, as has been the case in past times. (I rather thought the Japanese and Koreans had gotten over this practice, but maybe not. And if it's US companies storing nuts for the winter, then I'm really disappointed. <sigh> )
The statement "The increases will be just from incremental shrinks" hides a lot of sins as well. One form of incremental shrink, although we don't think of it as such, is the changeover to next generation (64M to 128M, for example). This doesn't always require a design rule shift. Some incremental shifts can change the die count from 205 to 545 on a single wafer, in one step; is that truly "incremental" when the percentage change is over 160? It doesn't take many DRAM superfabs performing this kind of shrink to get us into overcapacity with existing tool sets! Although I must say, I don't know of any pending, industry-wide shrinks of this magnitude. Seems like DRAM makers are indeed looking at smaller adjustments, such as narrower streets, slightly smaller die, and a few other technical tricks as they bulk up elsewhere for a real glut of overcapacity involving new 200 mm and 300 mm fabs.
There are still a large number of fabs that have not converted to 180 nm rules, though; and the gains available there are not to be discounted. Which leads to--
"She noted that lead times for delivery of critical production tools, such as wafer steppers, have stretched out to nine months." This may be the real fly in the soup. I think nine months is shorter than reality at present, especially for steppers that can write 180 nm lines. In some cases, that's all that is holding a line or fab back from the conversion from 250 nm lines. So they'll have to wait, even if they're ready with the rest of the tools. That's in addition to waiting for a Whole Fab, if that's the plan to get more capacity.
Ain't ramps fun? <g>
I think it remains to be seen whether there will be a major shortage, or if the industry will produce at better rates (including the release of any mythical stockpiles). My personal opinion is that we'll see DRAM go up to about $26-27 Billion this year, and on to maybe $31-32 Billion in 2002. The second number could be larger if Samsung really presses on their new 512Mbit chips beginning later this year. Those babies could haul down some serious coin, for a while....
Mitch |
