>>Isn't this a bit of a surprise? I thought that the drive to save costs was currently
focused on shrinking feature sizes. This would make that area the growth segment.
If there's not strong demand growth here then, by implication, chip companies are
not spending tons of money on new designs and shrinks. And the shrinks were
supposed to drive a lot of semi-equip purchases. Hardly what one would expect in
a recovery, U or V shaped. Am I overlooking something?<<
The picture is a little more complicated. First, and I suspect most important, ASICs and other custom circuits account for the majority of all masks. These customers don't generally have a design requirement for the higher density that 0.18 micron gives, or the design expertise to use all of those transistors. Given that, why should they pay for 0.18 micron masks? The foundries are actually facing a technology glut for exactly this reason.
news.semiconductoronline.com
Second, to hear the analysts talk, you would expect to see every chip company out there suddenly ditch their old fabs and buy all new equipment to support shrinks. This simply isn't true, and no one close to the technology has ever suggested that it was. Shrinks have been accelerating precisely because they *don't* require massive new equipment purchases. That's why the recovery to this point has been U-shaped.
news.semiconductoronline.com
Now the good news, and the reason why the Etec report isn't incompatible with a recovery, is that there's another scenario. Using advanced masks to shrink is less expensive in absolute terms than buying new equipment, but more expensive on a per wafer basis (unless you make DRAMs and use the same mask forever). So, if you expect your volume to support lots of new equipment purchases, that's what you do, bringing your mask costs down.
news.semiconductoronline.com
I don't think this third scenario is happening on a large scale, because the next generation of exposure tools (193 nm) just isn't mature enough to support large scale manufacturing yet, and because I'm not seeing the volume of orders that would indicate significant new fab construction. However, having 193 nm on the horizon might cause Etec's customers to delay massive mask capacity purchases.
(It's also a little more complicated because 193 nm for 0.18 micron (or, more likely, 0.13 micron) is still sub-wavelength lithography and will still require some mask tricks. Just not as many of them as 248 nm does at that feature size.)
Katherine |
