Back to the garage: Entrepeneurship lives on

01/01/1997

Back to the garage: Entrepreneurship lives on

In the early days of the semiconductor industry, there were no equipment vendors. Device makers pieced together process tools from component parts. Before integrated circuits, companies sold transistors and diodes directly to electronics OEMs, with very little concern for the eventual application. These components were simple and the designs were commonly known, so competitive advantage came from processing. Companies offering more consistent, reliable parts were more likely to succeed.

A lot has changed since then. Semiconductor companies expect equipment makers to provide complete systems for each process step. The electronic equipment OEMs buy chip sets for multimedia, communications, or general purpose computing, and the designs of these chip sets are closely guarded secrets. Market advantage comes from circuit designs, and from robust process flows linking pieces of equipment.

Yet, an advanced CPU and supporting circuits can require tens of millions of transistors, designed by huge engineering staffs. An advanced CMOS process can require 250 steps, performed in a fab costing nearly $2 billion. The enormous combined overhead of design and fabrication shuts out all but the largest companies.

Enter foundries and fabless design houses. All a foundry does is make chips, devoting all of its engineering expertise to fabrication. Market success requires advanced processes and fast turnaround times, and profitability demands increased flexibility and reduced changeover time between designs. Meanwhile, fabless semiconductor companies only design circuits. They give up control over the process in exchange for increased flexibility and reduced overhead, and succeed by offering specialized products, tailored to specific applications and produced in relatively small quantities. Foundries thus re-open the market for small, entrepreneurial designers who could never afford a fab.

Foundries are not going to take over IC fabrication, though. As many fabless companies have found recently, a fab with more than one customer may not be able to increase a particular customer`s production quickly. As more and more consumer products incorporate semiconductors, "standard" chip sets, like those used in personal computers, will emerge for home automation, automotive guidance systems, personal digital assistants, and other mass market applications. These applications will still demand the huge quantities that only captive, single-purpose fabs can provide.

Chip companies with captive fabs are starting to concentrate on circuit designs as well, asking equipment makers to shoulder as much of the processing burden as possible. Already, equipment companies are absorbing most of the cost of 300-mm tool development. Some of the larger vendors now offer integrated, multistep processes to produce complete structures on the wafer. Further down the road, several equipment company CEOs say they expect to see equipment acquisition shift from traditional purchasing to a model in which the equipment vendor provides a complete process cell and is paid by the wafer.

In all of these scenarios, the chipmaker gives up control over the process in exchange for simplicity. Expertise shifts from the chipmaker to the equipment maker, who must reduce cost/good wafer in order to compete. The large research and development investments required by this model threaten to increase the dominance of large equipment companies.

Still, there is room for equipment entrepreneurs, too. As large equipment companies become integrated process suppliers, they will look for more innovative component vendors to help them shave costs and improve performance, and high-performance software to tie the pieces together.

This issue of Solid State Technology marks the debut of our new AsiaFocus column. This department, which appears on page 40, offers a closer look at events and trends in the Asia/Pacific region.