For wireless apps, a specialized chip application, it's like GaAs.
Overall, it's better than GaAs and allows for integration of digital functions on the same chip.
For Intel's (or IBM's) main output of wafers, it's not going to get integrated-in anytime over the next several years.
I will guess that IBM will do more than $1 billion in foundry work on Si/Ge and Si/Ge BiCMOS OEM designs in 2002. This business is increasing at over 20% a year.
The same analysis could have been made about many niche technologies (SOS, MBE, I^2L, etc.) back in 1980.
Si/Ge and Si/Ge BICMOS is not a niche technology.
I'm an investor in a wireless company doing multi-GHz/ultrawideband chips. Their main need is for processing speed, at high integration densities. Pure blazing transistor speed is only need at the front-end (near the antenna), not at the back-end. (Sure, they'd take the pure blazing speed at the back-end if it cost the same as 0.18 micron CMOS, but it doesn't, and won't for many years.)
You really don't get it. With Si/Ge BiCMOS, you get both on the same chip. The SiGe Bipolar gives you the blazing speed for the front end and the conventional CMOS (without any
performance or density loss)is used for the back end functions. Given the way the integration is done, it is not that much more expensive than alternative techniques. I believe they currently have a integrated .18um BiCMOS process with a .13um process well on the way. You really have not kept up to date on recent advances. Even a cursory
examination of recent information on Si/Ge and Si/Ge BiCMOS on the web will show how dated your views are. I expect Intel to try to move quickly into this area.
THE WATSONYOUTH |
