The Changing Path of Taiwan's IC Industry

Carol Liao, Editor, Taiwan -- Semiconductor International, 10/1/1998

Confronting the downturn

Taiwan's IC fabrication business fluctuated like a roller coaster in recent years. In 1997, total revenue of the country's IC manufacturing business was $5.3 billion, with an annual growth rate of nearly 20%. Ninety-three percent of that was from memory and foundry business. Though 8 in. wafer capacities are expected to expand in 1998, and DRAM pricing may slide downward in a more stable manner, it appears as if the market will stay in a slump until the end of 1998. Table 1 is a modified forecast of Taiwan's 1998 IC industry revenue from ERSO ITIS. As can be seen, every section of the Taiwan IC industry is influenced severely by financial turmoil in Asia and the general industry downturn.

"What caused the industry recession is not only the Asian financial turmoil but also the popularity of low-cost computers as well as loss of confidence," said Chiam Wu, president of Applied Materials Taiwan and corporate vice president of Applied Materials. "Compared to other countries, Taiwan still stands in a better position, because its foundation is done well, with efficient throughput and talented human resources."

Applied Materials, the largest semiconductor equipment supplier, had closed its Korean office and downsized its office in Japan to pass through the slump, but strengthened the importance of its "Technical Center" located in Taiwan.

Strategic alliances common

Taiwanese semiconductor vendors are looking for ways to accumulate technology capability. Over the past two years, the number of international strategic alliances within the Taiwanese semiconductor industry has increased continuously, and these are now common practice.

TSMC and UMC are the two largest wafer foundries in Taiwan. Following the recent slump in the DRAM market, DRAM makers ­ including Acer (formerly TI-Acer), Powerchip, WSMC and Nanya ­ also have moved into subcontract wafer manufacturing. While the effect on the two manufacturing "giants" is not expected to be too great, it could bring pricing chaos.

As the first and largest foundry company in the world, TSMC achieved much in sales and technology. In 1997, TSMC announced it would invest more than NT$400 billion (U.S. $11.59 billion) (UMC issued NT$500 billion, U.S. $14.49 billion) over the next 10 years to build six advanced fabs in the Tainan Science-Based Industrial Park. Meanwhile, WaferTech, its joint-venture fab in the United States, has begun volume production also.

In 1996, TSMC formulated the company's vision of the "Virtual Fab" to develop long-term competitive advantages by strengthening the bond with its customers. In early 1998, it completed development of 0.25 µm process technology and began 0.25 µm volume production. In addition, TSMC is expected to be the only company in Taiwan scheduled to start 0.18 µm process technology in the third quarter of 1999.

Strategically allied foundry group: UMC Group

In 1995, UMC, Taiwan's first wafer fab, invited 11 leading IC design firms from North America to establish three new 8 in. wafer foundry companies jointly. In 1997, the UMC Group completed its transition program to join the foundry business. Then in 1998, UTEK joined the UMC Group and became the fifth foundry company.

The UMC Group could be regarded as the model of
cooperation for establishing wafer fabs in Taiwan's IC industry. Cooperation is one reason it has been able to keep pace with leading international companies in rapidly ramping up 0.25 µm process technology. The UMC Group has engaged in research and development of 0.18 and 0.13 µm (and beyond) process technology to meet future foundry demand.

Powerchip & Mitsubishi

Powerchip Semiconductor Co. has maintained a cooperative business relationship with Mitsubishi. From early 1998, Mitsubishi has devoted much effort to producing mainstream DRAM products as well as special value-added goods. At present, the 64M DRAM is produced only at Mitsubishi's fab in Japan and at Powerchip in Taiwan.

Mitsubishi has handed production of its basic mainstream products to Powerchip but still controls manufacturing of high-end special application memory. Powerchip's trial production of the 0.25 µm 64M DRAM began in late 1997, and it was mid-1998 before the company began construction of 0.2 and 0.18 µm vendors.

Mosel-Vitelic & Siemens

Mosel-Vitelic and Germany's Siemens jointly established ProMos Technologies, utilizing Siemens' 64 and 128M DRAM technology to develop products. Currently, ProMos Technologies produces both 0.35 and 0.25 µm 64M DRAM. The technology to produce 0.18 µm DRAMs probably will be developed by ProMos Technologies in 2000.

Acer vs. TI

In the past, TI-Acer relied on Texas Instruments for technology. This, however, is in transition. TI-Acer is gradually reducing production levels for DRAM products in order to dilute management risk. As a result, TI plans to gradually wind up its cooperative relationship with Hitachi and focus primarily on digital signal process (DSP). In April, TI diverted holdings but agreed to allow TI-Acer to continue production of the 64M DRAM for another year. The entire production yield will be sold back to TI.

Cooperation with TI has allowed Acer Semiconductor to master special application memory design technology that will benefit its future transition to a multi-function wafer company.

Nanya & Oki, Winbond & Toshiba

Other important technology transfers underway are those between Mosel and Oki, Nanya and Oki and Winbond and Toshiba. Because Winbond had no previous experience in DRAM technology and wanted to enter the DRAM market, the company initiated a technology joint venture with Toshiba. They first utilized a DRAM foundry to obtain the technology. Approximately 50% of the initial production yield will be returned to Toshiba.

Macronix & Matsushita

Macronix also transferred DRAM production technology from Matsushita. In this way, Macronix was able to enter into the realm of embedded flash and interior-embedded DRAM manufacturing. Since the development of special application memory products is moving towards embedded DRAM technology, the first contract for the transfer of technology signed between Macronix and Matsushita states that the 16M DRAM plan cannot exceed 1/3 of the total yield from Macronix's two factories. In addition, with assistance from Matsushita, Macronix will extend production technology to 0.25 µm 64M DRAMs.

WSMC & Toshiba

The technology cooperation agreement between WSMC and Toshiba, one of the world's 10 largest semiconductor companies, is another example. WSMC was established by a partnership between China Investment Corp. and Winbond. Currently plans are primarily for the production of a 0.25 µm process but also will include a 0.18 µm process in the future. They were forecasted to begin operation in July and will begin volume production for their foundry business by the fourth quarter. Table 2 lists Taiwan's top 10 IC vendors.

Rooting Taiwan's advanced technology capability

Over 60% of Taiwan's semiconductor companies already have developed 0.25 µm technology. Currently, ITRI is engaged in a deep submicron project to develop next-generation 0.18 µm technology. This plan does not seek to develop a total technology package; rather it looks to develop key module technology, leaving remaining technologies to be developed by individual companies.

"Taiwan's IC industry is increasing in importance, but more investment than performance," said Genda J. Hu, vice president and ERSO general director of ITRI. "The island clustered 18 8 in. fabrication factories together that are all good at manufacturing as second source. Now it's time, however, for Taiwan's IC vendors to transform their strategy from investment-oriented into innovation-oriented, so as to keep in step with leading countries like the United States or Japan." As Taiwan's industry moves into its next major evolutionary phase, 300 mm wafer technology, TSMC is the only company in Taiwan that has joined SEMATECH to participate in the I300I project.

The concept of the Center for Semiconductor Research is similar to those consortia led by members of the industry such as SEMATECH in the United States, Japan's ASET and Europe's IMEC, targeting very advanced semiconductor process technology and materials as its founding objective. "The purposes of 'Taiwan SEMATECH' are to cultivate Taiwan's long-term R&D capability, to balance the allocation in R&D resources and to strengthen competitiveness, upgrading the R&D efficiency," Hu said. "The mission is to develop pre-competitive core process technologies, like copper interconnect and low-k processes, to evaluate new process equipment and to assist the forming of semiconductor peripheral industries."

"The R&D team of the 'Center for semiconductor research' will come from ERSO's deep submicron project in the beginning," Hu said. "We are looking forward to setting up the consortium in the end of 1999; until now over 10 IC fabrication vendors expressed their willingness to join in. We look forward to all Taiwanese semiconductor makers joining the technical center to work with worldwide equipment vendors, promoting their technology capabilities. Taiwan SEMATECH wants to keep partnership with worldwide equipment vendors, not to impose restrictions on ourselves."

"Government should cultivate talented manpower to match the industry's demand in time," Wu said. "Besides, Taiwanese wafer vendors should pay more attention to avoid ESH events, such as the fires at Winbond and UICC, which resulted in dual losses of huge assets and company image."

Conclusion

Even though Taiwan's semiconductor industry is able to compete with the rest of the world when it comes to manufacturing process technology, Taiwan still needs to strengthen its product technology. Taiwan also has room to improve on design, innovation and application. The tasks of demonstrating its formidable production power, strengthening technological capacity as well as lengthening the distance from new competitors is likely to be the focus of Taiwan's semiconductor industry at this stage.