Foundries flourished largely because of the continued disaggregation of IC design and manufacturing, which has its roots in the formation of third-party providers of libraries, intellectual property cores, design tools and expertise. The creation of the foundry business by TSMC chairman Morris Chang in 1987 was the catalyst for this trend, allowing small companies to build entire product portfolios without any in-house semiconductor manufacturing capability. It expanded innovation

and established the disaggregated foundry model as the most important IC business model in years. What's more, in the decade since the foundry model became a viable force, no new integrated device manufacturers have been created.

As foundry proficiency grows, new capabilities emerge. TSMC can produce more than 4 million wafers a year in leading-edge 0.15- and 0.13-micron processes on 200-mm or 300-mm wafers. Foundries like TSMC also have the broadest semiconductor technology, including CMOS logic, mixed-signal and RF, embedded memory, image sensors and specialty technologies soon to be introduced.

We believe 2001 will be the year of the system-on-chip (SoC). Indeed, companies have claimed SoC capabilities for years. In truth, they probably do-based on earlier requirements for chip-level systems. But this year will mark a milestone. For the first time IC companies will offer integrated logic, mixed signal, RF, memory, CPU and peripheral circuitry on a single chip, supplied by a foundry, with the power range, performance and price point necessary to enable innovation.

The challenge is to develop these systems in a disaggregated environment. Where is the methodology for SoC designs? Where are the Spice models, the libraries, the IP and the standards for evaluating all of these SoC components? And how do you validate a mixed-signal design on deep-submicron technology and still get to market first?

That is the key challenge. Are found-ries prepared for it?

Since foundries recombine the resources of both customers and IC service providers, they are uniquely positioned to participate with "the best of the best," working collaboratively to solve advanced process and design issues. From these efforts arise new memory cells, libraries and IP blocks often developed at TSMC first in return for validation through our CyberShuttle program.

Foundries have contributed significantly in the past 10 years. We believe they will lead the way to high-volume production of exciting new applications employing true SoC technology.

Edward C. Ross is President of TSMC North America (San Jose, Calif.).