Notes on semiconductor technology by D.Messerschmitt

(based in part on the lecture by Chenming Hu)

The basic message of this talk is that standard CMOS technology has been following an exponential curve by any cost/performance measure. This advance can be expected to continue for at least one or two more decades. There are several factors that could ultimately limit the technology:

• Fundamental physical limits.
• The manufacturing capital investments become so large that sufficient investments cannot be attracted. This seems unlikely if we consider that there are about 40 latest-generation factories, which is much smaller than the number of electric generating plants. Capital investments have recently been accelerating.
• The applications do not demand higher performance. For example, supersonic passenger plans are technically feasible, but the marketplace does not demand them.
• Consumers become unwilling to spend more of their income on electronics. Chenming predicted that in fact the electronics share of the gross world product would saturate at some level, looking at the history of other industries like the automobile.

For the time being, there does not seem to be any obstacle to further advance in semiconductors on the horizon. This has significant implications applications dependent on this technology, like computing, networking, etc.

Another interesting issue is whether new technologies that compete with CMOS might be able to gain a foothold. Even if a new technology is inherently superior, it may not be able to gain a foothold against CMOS unless:

• It can overcome the huge accumulated investment in CMOS technology. This investment includes not only the direct investment in the design of CMOS processes, but also the investment in complementary technologies like semiconductor manufacturing equipment, computer-aided design software, training of the current generation of designers, packaging, etc. Thus, the current marketplace is strongly influenced by past investments. This is a path-dependent effect, where the past history influences the ability of the current market to sort out the best technology.
• Related to this is the cost to customers of CMOS technology to switch to another technology. These switching costs create a lock-in effect to CMOS technology.
• However, we should also recognize that in a growing industry, where new factories are being built, there is a greater opportunity to introduce new technology without divesting the current technology. Thus, this situation is different from the lock-in of the RBOCs to their switch vendors, since the number of switches is not growing. Also, some of the investment in CMOS is probably directly transferable to other related technologies like GaAs, for example in lithography.

The bottom line is that a new semiconductor technology has to be a really significant advance to displace the current technology. When technologies are competing that have relatively undifferentiated characteristics, like GaAs and Silicon, then the technology that attracts the greatest investment will be the winner. It will also have a strong tendency to take over 100% market share because of economies of scale, such as the spreading of the cost of development of CMOS-related CAD software and manufacturing equipment across a larger industry base.

Thus, the CMOS technology illustrates at least three economics effects:

• Economies of scale in manufacturing and supporting industries
• Switching cost and lock-in to the dominant technology
• Path dependence on accumulated investments in the dominant technology

Copyright 1997, Regents of the University of California. All rights reserved.