Taiwans future linked to foundries
02/01/1998
Taiwan`s future linked to foundries
Alan Jung, SEMI, Taipei, Taiwan
The balance of power in the semiconductor industry is set to shift as foundries become a key driver for wafer processing technology and the IC design community sees the emergence of "chipless" companies driven by intellectual property development.
These and other issues were raised and discussed during the inaugural Taiwan Industry Strategy Symposium (ISS) 1997, co-sponsored by SEMI and the Taiwan Semiconductor Industry Association (TSIA).
In the keynote dinner address at Taiwan ISS 97, W.J. Sanders III, chairman and CEO of Advanced Micro Devices Inc., said the major challenge facing a Taiwan semiconductor maker today is to build a long-term viable enterprise through participation in value-added products. Such products provide a higher level of insulation from the vagaries of commodity markets, offering a better opportunity to achieve an acceptable return on investment. Sanders said there were two ways to participate in value-added products: as a foundry service or as an integrated device manufacturer (IDM) developing an in-house product portfolio.
James Hines, principal analyst for Dataquest Inc., said that despite the slump in DRAM prices, growth in memory bit demand is expected to outpace price declines in 1998. However, Hines warned that lower cost structures due to currency devaluations in Asia could see slower revenue growth in DRAMs this year.
By 2000, the total semiconductor market will be worth $272 billion, up from $175 billion in 1998, according to Dataquest. Total capital spending for chip manufacturing will reach $44 billion in 1998 and $77.5 billion by 2000, Hines said.
Foundry models
In the 1980s, DRAMs were the technology driver for wafer fabs. In the 1990s, the driver has shifted to be CPUs. As we move toward the next century, ASIC and logic devices will become key drivers in semiconductor technology. Fabless companies are at the forefront of combining memory and logic. "This puts the foundry at the center of the semiconductor industry. We see the fundamentals in place for a very solid foundry growth," said Hines.
|
In 1997, foundry work represented about 7-9% of worldwide semiconductor production. Dataquest expects that to rise by 2001 to 14 or 15%, and by 2010 the share of foundries is projected to reach between 35 and 45%.
At the same time, foundries are starting to provide greater levels of design service. In the future, foundry designs will start to encroach on traditional ASIC markets, forcing companies in these markets to rethink their business models in response to the increasing competitive pressure from foundries.
Another speaker who addressed the foundry issue was Morris Chang, chairman of Taiwan Semiconductor Manufacturing Co. (TSMC). Although the future looks bright for foundry companies, Chang highlighted some challenges to be overcome before the 35-45% foundry market share threshold could be reached. "To get to the 40% number we will probably need a few of the large IDMs to join the ranks of our customers," he said.
Firstly, Chang said, the costs of using a foundry were not well understood by IDMs. A "virtual" foundry such as TSMC has a lot of overhead, so the total cost of a foundry cannot be compared with the manufacturing cost of an IDM.
A second problem is that of company politics in the IDM. "The fab people in the IDM would like to build more fabs because they grew up in a fab environment," says Chang. Typically, only about 20-30% of the IDMs` wafer fab needs are outsourced to the foundry, while the rest is kept in house.
|
Figure 1. Taiwan ICproduct revenue trend. Addition of figures in first two blocks of each solid column yields numbers on top; addition of all three blocks in each solid column yields numbers on bottom.
Another perspective on foundries was given by Peter Chang, president of United Semiconductor Corp. He predicted that in 10 years time there will be fewer part-time players - IDMs offering part of their production to third parties. Oversupply in the foundry market means customers will choose dedicated foundries. "The foundry business is relatively lucrative, so it is attracting new players," he said.
Ding Yuan Yang, president of Winbond Electronics Corp., provided a perspective from the IDM. He pointed out that the high cost of new wafer fabs has created a barrier to entry. As a result, since the early 1990s, there have been no new IDM start-ups. The need for low-cost manufacturing, high-volume production, and flexible delivery of product matched the strengths offered by IDMs in Taiwan.
There is, though, a downside to being an IDM, according to Yang. An IDM has to grow very quickly to reach economies of scale in order to overcome market entry barriers. IDMs must also maintain high levels of R&D spending, while also investing heavily in management information systems (MIS) to handle the manufacturing and logistics of a large portfolio of products in a volatile market. Finally, IDMs must take time to build up a portfolio of intellectual property to fend off litigation.
Frank Robertson, of the International 300mm Initiative (I300I), gave an outline of the progress toward 300-mm fabs. He said that it used to be possible to make smaller increments in wafer size, such as a 33% increase from 6- to 8-inch wafers. "Now we have to have a 50% size increase to justify the cost of conversion," he said. By the second half of 1998, several 300-mm pilot lines will be up and running, typically producing 2000-5000 wafer starts/ month. The first pilot production line was installed in Q4 1997. Robertson believes that by 1999 and 2000 the industry will start to see serious ramp up in 300-mm production; by 2001, virtually all new facilities will be 300-mm.
In Japan, the shift to 300-mm will also signal a shift to minienvironments. While Taiwan fabs have been heavy users of minienvironments, the Japanese have preferred the open cleanroom concept. Now, eight out of 10 J300 member companies say they will use minienvironments for 300-mm production.
Adding value
Nicky C.C. Lu, chairman of Etron Technology Inc., discussed the concept of the intelligence-oriented firm. These so-called "chipless" companies are an extension of the fabless concept - with the emphasis on development of intellectual property (see table). Chipless companies are evident in the US, but none has yet emerged in Taiwan. Lu said that Taiwan engineers preferred to work for a manufacturing company, not an intelligence-driven firm. In the US, the opposite was true. "Taiwan is a manufacturing-driven model. That is important, but it is not the only way to win," he said. Lu urged more Taiwan companies to adopt the intelligence-driven approach (Fig. 1).
Echoing similar sentiments was Paul Franklin, senior VP of operations for S3 Inc., who noted that the emergence of state-of-the-art foundries like TSMC has made leading-edge IC technology available to anyone who wants it. Process nowadays is more defined by the equipment suppliers, according to Franklin. This has leveled the playing field between the IDMs and the foundries. Franklin also agreed that software and "intelligence" applications will be the key differentiator in the future. One application that will drive silicon integration and consume larger amounts of silicon is 3D, which requires over two million gates to implement. "3D will eat up gates left and right for the next five years," predicted Franklin.
In a keynote address, Matthew Miau, chairman of Mitac Inc., pointed out that it was becoming difficult to distinguish between manufacturers and distributors in the PC world. For example, one major US computer distributor introduced a build-to-order system called "channel assembly" in order to cope with ever-shortening PC product life cycles. Such changes also impact the semiconductor industry, with wafer fabs and IC packaging operations looking at new business models to match the changing PC distribution model.
Lower component prices (i.e., DRAMs and TFT displays) and increased silicon integration will allow system makers to devise affordable, all-in-one consumer products. For example, Taiwanese companies are now looking carefully at how to integrate the PC, TV, set-top box, Internet box, HiFi and DVD all in one unit. For the office market, a combined PC, fax, scanner, copier, and imaging system is envisaged. "This provides a lot of opportunity to design and provide chips that integrate these functions," said Miau.
|
Figure 2. Year-to-date (Jan.-Sept.) orders and shipments by region.
Elizabeth Schumann, an analyst with the SEMI market statistics group, told Taiwan ISS 97 that 1997 would be a "slow but sure recovery" in semiconductors. However, since an estimated 30% of all PCs sold in 1997 were sub-$1000 machines, unit sales were up, but average selling prices were down. The bottom line? Lower margins for device makers and tightening capital equipment budgets.
Despite the slowdown in the industry and low DRAM prices, semiconductor equipment sales continue to grow (Fig. 2). "Capacity needs and new technology needs are driving new equipment sales," she said. "The equipment industry is now out of the slump and heading for the upcycle." According to SEMI`s November 1997 consensus forecast, the CAGR for worldwide semiconductor equipment revenues will be 17% over the three-year period from 1998 to 2000.
The following trends can be identified from SEMI`s latest market statistics:
The Japanese market for equipment contracted by about 10-15% in 1997. This was due to three factors: the weakening of the yen, reduced capital expenditures, and the building of more Japanese fabs offshore.
The US has regained its position as the leading manufacturing location for semiconductors, with total capital equipment orders for 1997 (January to September) reaching US$7453 million.
Taiwan has become the third largest market for capital equipment, behind the US and Japan, and ahead of Korea and Europe. For the nine months to September 1997, total orders were US$3257 million.
Despite the widespread belief that there is overcapacity in the device manufacturing industry, SICAS (Semiconductor Industry Capacity Statistics) data shows that capacity utilization for high-end devices is still tight. For example, in 2Q97 (the latest period available), MOS wafer fab utilization at <0.4 ?m (6-inch wafer equivalents) was 94.6%, up from 89.5% utilization in 1Q97.
Globalization
Karlheinz Horninger, executive VP of ProMOS Technologies Inc., said the globalization of the semiconductor industry has forced IC makers, their equipment suppliers, and the chip-buying customers to adapt to new rules and conditions.
For the equipment and materials vendors, globalization means they must follow their customers and set up service and maintenance centers close to the wafer fabs. In addition, it may also be an advantage to have equipment and materials manufacturing capabilities in different regions of the world.
In the IC manufacturing community, there has been a proliferation of front-end and back-end facilities sites around the globe, providing a high level of flexibility to the semiconductor makers. Horninger said the challenge has been to convince customers of the advantages. Product delivery from different locations of one semiconductor manufacturer is still difficult and time-consuming, since multiple site qualifications are required.
Horninger said the "fab cluster" concept was a new approach to help solve this dilemma by ensuring equal quality from all the front-end and back-end sites of the supplier. In a fab cluster, all the front-ends are synchronized in such a way that a high level of physical and electrical equivalence is achieved. This leads to the same level of electrical and mechanical quality of parts, so that the customer can receive deliveries of parts from different manufacturing sites without tedious and costly multiple qualification procedures.
Conclusion
In order to remain competitive in the future, Taiwan`s semiconductor industry must undergo a shift from a manufacturing business model to one more focused on adding value. Globalization means that device makers, their customers, and the equipment and materials suppliers must adapt to new conditions and forge alliances with one another. Although the semiconductor industry has gone through a correction, the outlook for 1998 appears to be one of "cautious optimism," especially for device makers in Taiwan.
ALAN JUNG received his BS degree in electronics and computer science from San Jose State University and his MBA degree in finance from California State University, Hayward. He has been the SEMI regional director of Southeast Asia, based in Hsinchu, Taiwan, since November 1995. Prior to relocating to Taiwan, he was the manager of the Technology Group at SEMI headquarters. SEMI Taiwan, 14F-1, Empire Commercial Bldng., No. 295, Sec. 2, Kuang Fu Rd., Hsinchu, Taiwan 300; ph 886/3-573-3399, fax 886/3-573-3355, e-mail, [email protected].