Andrew and threaders,
It's my understanding, from the INFRASTRUCTURE talk I attended, that there is more than 30 fabs (maybe as many as 36) coming online this year and next, meaning already under construction, yes? In any case, that's not the nit I wanted to pick. <g>
The talk I went to, the data shown suggested that 300mm and 200mm new fabs were about equal numbers. And since the 300mm fabs were planning fewer wafer starts per month, the actual chip output from each fab wouldn't be that much different. For now, of course! Once the 300mm folks figger out their processes, they'll ramp them on up to as many starts as they can, just like everybody has done in the history of this fine industry.
That's like printing money at that point.
I don't think these fabs will have near the impact vs. demand in late 2001 that they will have in late 2002, when they a) begin to reach higher wafer throughputs, while b) reaching their third generation of product, predicted to be about 2.5 times as many die per wafer as their startup product.
That's a pretty typical profile for 200mm superfabs as well, where they are loading about 2x the wafers at full production that they did when their first product hit the street. And some get as many as 3x die counts within 1.5-2 years.
Which means one 300mm fab doing, say, SDRAM, can expect to produce about 5% of the world's chips in that product class.
Even with growing demand, we don't need too many of these monsters! So I now think we'll see us reach the overcapacity point in late 2002 or early 2003, and it'll be a Big One on this overshoot.
That's even with ramping demand for chips of all kinds....
Mitch (Just my opinion) |
