Editorial

The Foundry Business Is Booming, but Will the Good Times Last? Cheap silicon attracts new customers, but system-level integration and a likely foundry glut cloud the future. by Jonah McLeod

In September, I traveled to Taiwan to determine what was happening in the battle between the ASIC vendors and the foundries. Ted Tsai, general manager of Maojet Technology Corp., a major rep firm in Taipei, and one of his U.S. customers, Chungsen (Charlie) Chen, president of Lexra Inc., a start-up intellectual property company based in Cambridge, Mass., arranged my meetings in Science-Based Industrial Park in Hsinchu.

My visit with Ming-Kai Tsai (no relation to Ted), president of the Second Business Group of United Microelectronics Corp. (UMC), enlightened me about the origins of most foundry designs. The largest number of designs coming into UMC are still destined for the PC in the rapid drive to add 3-D graphics and motion video to the PC platform.

He says, however, that foundries are also attracting new industry segments, notably telecommunications equipment vendors. As with PCs, cellular phones and other telecommunications equipment contain a CPU, memory, glue logic, and mixed-signal circuits. The low cost of silicon for the glue logic and mixed-signal circuits has attracted designers of ASICs for those products to the foundries.

A potential problem for the foundries, though, is that future generations of products will integrate the processor, memory, and glue logic all on one chip. Thus they could be forced out if they lack the processor and memory cores that designers need. Texas Instruments and Motorola aren't likely to license their DSP and microcontroller cores to a foundry.

But third-party foundry-independent processor cores, like the ARM and MIPS cores, are bolstering the foundries. In addition, foundries such as UMC are arming themselves by developing strategies to offer designers embedded memory as well.

But designers wanting access to low-cost silicon aren't the only allies the foundries have. In my discussions with Peter Chang, president of United Semiconductor Corp. (USC), a spin-off of UMC, I learned that major semiconductor companies making standard products see the foundries as their partners. The reasons are the foundries' ability to provide state-of-the-art process technology and the high cost of building new fabs, which is well over $1 billion.

Chang says that IC companies build less capacity than they anticipate they'll need and rely on the availability of foundry capacity to supply the shortfall. Just in case there's some question about the economic viability of the foundry business, Chang reports that Taiwan foundries achieved after-tax profits of better than 30 percent--a figure comparable to Intel's. And the forecast remains bullish for the future. Chang points to a recent rise in price at USC rival Taiwan Semiconductor Manufacturing Company, also in Hsinchu.

Such attractive profits are encouraging a boom for new foundries in Taiwan. M.K. Tsai knows of at least 10 new 0.35-µm and smaller-geometry fabs that are either under construction or currently planned. Such a build-up is taxing the water and power infrastructure of Hsinchu. Therefore much of the new construction is going on at the southern end of Taiwan, around Kaohsiung.

The pace of new construction worries individual foundry companies, such as UMC. M.K. Tsai cites a forecast from the San Jose-based market research firm Dataquest that a glut in capacity is likely to occur when many of the new fabs come on line before the end of the decade.

In the meantime, there's a lot of money to be made in the pure foundry business. *

To voice an opinion on this or any Integrated System Design article, please e-mail your message to miker@isdmag.com.

integrated system design  November 1997