August - 2000 - issue > Cover Feature

The semiconductor industry has never been able to rid itself of cyclical whims of its consumers. Sales typically go up and down with a periodicity that baffles even the most ardent students of chips. Most analysts say that we are currently in the ninth phase, which started last year. The previous phase started in 1992, and reached a climax in 1995, before grinding to a halt by 1998. Part of the uncertainty lies in the time it takes to build a chip fabrication unit.

Sudeep Balain of Chase Securities, an industry analyst who spent 10 years in the industry and came to Chase via Wit Soundview, argues that “if a semiconductor company today wants to pull the trigger on a new fab, it will take them at least 21 months to get from ground zero to a run rate of about 1,000 wafer starts per week.” These 21 months can be broken down as follows: seven months to build the shell or structure, the next seven months to equip the fab and build the so-called “clean room,” and the final seven months to install various kinds of equipment and run reliability tests.

The importance of the structural design of the fab can hardly be understated. In the case of Intel, for instance, it transformed the chip maker from a 80-pound weakling into an 800-pound gorilla in manufacturing by mastering a technique billed as “Copy Exactly.” Intel was able to design fabs faster than anyone else by creating twin plants identical in every respect down to the color of the paint on the walls. By doing so, it was able to emulate production technologies and boost scale in all plants simultaneously. (Intel is currently the largest semiconductor manufacturer worldwide, with 20 fabs that cumulatively have a capacity of 437,000 wafers a month.)


PC No Longer King

There are some fundamental changes rippling through the semiconductor industry; because of it, all the incumbent firms from Intel to AMD to IBM are struggling to find meaning and clarity. As recently as 1995, 50 percent of chips produced went into PCs. That is no longer the case. In 1999, that figure has reached the 25 percent range and is rapidly declining. However, it is not that the number of parts in PCs are growing smaller, it is that the number of other devices capable of accessing the Internet has exploded.

Worldwide semiconductor sales increased by 40 percent in May from a year earlier, sending the Semiconductor Industry Association into raptures. The recent growth will keep the industry on track to produce at least a 30 percent rise in production during the span of the year. Internationally, sales grew 46 percent in the Asia Pacific, 44 percent in Japan, 33 percent in the Americas and 40 percent in Europe.

So the question on everyone’s mind is: why is this sudden spurt of growth happening right now?


Industry Drivers

In a word, the Internet. The industry has shifted away from being PC-centric to being Internet-centric — or, according to the enthusiastic Balain, “new economy-centric.” This shift makes the chip manufacturers much less susceptible to wide swings in demand or competitive entrants that blindside a corporate behemoth. A classic example of the latter is when Compaq introduced the $999 PC using a Pentium clone in February 1997. Soon after that event, Intel’s share of the low end dropped to about 30 percent.

A second reason is the evolution of converged networks. People are rapidly standardizing across various platforms, which will help businesses improve efficiency by a scale multiple rather than mere percentages. Also, as B2B and B2C continue to grow, these Internet transactions demand efficient networks to travel along. And of course, the pulsing heart behind all of this is that tiny chip.

A third reason is again related to the Internet: the number of devices, and the kinds of devices that can communicate with each other are rapidly proliferating and shamelessly intruding into every facet of life. The professional with a cell phone has come to symbolize the present age. Sub-$300 devices are sprouting up all over the home ranging from Sega machines to Sony’s robotic dog, currently being tested in Japan. In a bizarre twist, fuzzy logic has been used to make the dog idiosyncratic and unpredictable: it is designed to bark and snap and be fussy.

Currently, however, given the humungous growth in the semiconductor industry, the metaphor of a snappy little dog is not the appropriate one to sign off with. Chips are glamorously efficient. Industry analysts are rationally exuberant. Perhaps, just perhaps, those dreaded business cycles have come to an end. The industry has outgrown two-wheeled cycles, and proudly climbed into a new motor car. We can now wait in abeyance for the airplane age.


Krishnan Sethumadhavan works in strategic planning for a Connecticut-based startup.