In it for the long haul? GaAs foundries in Taiwan
11/01/2001
Andy Bavin, Trikon Technologies Inc., Ringland Way, Newport, UK
 |
The GaAs market is widely considered the fastest growing arena of semiconductor technology. During 2000, it experienced some 60% growth over 1999, with device shipments totaling more than $18 billion. Spurred on by the prospect of increasing demand for wireless communications and optoelectronic and LED lighting products, the sector seemed poised for accelerated growth at the end of last year.
The GaAs sector, however, has slowed dramatically along with the rest of the semiconductor market, as predicted by Dale Pfau of CIBC World Markets at a conference early this year. The proliferation of SiGe HBT technology is also undercutting market prospects as chipmakers strive to make communications devices using CMOS silicon rather than the more costly GaAs technology.
During the last 10 years, a number of companies have established GaAs foundries, serving as catalysts for the industry's evolution from specialist fabrication to large-scale mass production. Taiwan has become the center of this GaAs foundry revolution (see the table) because of its knowledge, skills, expertise, and success with silicon industry applications. The precipitous market downturn and threats from SiGe and InP technologies have now cast a shadow over the island's seemingly bright future. This article will examine Taiwan's growth in the GaAs industry and discuss whether current market conditions dictate using the wafer fab or the foundry model for continued GaAs production.
GaAs foundry model benefits
To stave off the threat of rival SiGe and InP technologies, GaAs ICs are being continuously reinvented, redesigned, and improved. Even as production volume slows, the continual evolution of new and better device designs proceeds unimpeded. Consequently, GaAs ICs often have a short operational life cycle.
Therefore, time to market is critical to the economic success of any new device.
Foundries have capitalized on this element of profitability. As job shops, they must produce small quantities of a large number of designs in a short period of time. Many have achieved this objective by basing their design and operation on high-productivity Si fabs, using high-volume process equipment. Although initial investment in such a project is often too high for low-volume GaAs fabs, foundries benefit in the longer term due to their additional functionality, extendibility, and productivity.
 |
Advanced manufacturing equipment not only facilitates process control, but also the ability to easily change and refine processes to meet constantly changing design requirements. These tools offer foundries the capability of manufacturing on 150mm substrates with high processing throughputs, increasing the volume of product that can be processed in a short time. With effective factory control systems, such as work-in-progress control and an effective management information system, foundries can reliably monitor their output, making possible an improvement in device yield and reducing material losses and their associated costs.
Taiwan's foundries
Despite huge differences in overall market size, the boom in Taiwan's GaAs foundry services was largely prompted by the successes of Si foundries such as TSMC and UMC. In 1991, Hexawave Photonic Systems became the first (albeit small-scale) GaAs foundry to be established on the island, opening its factory in Hsinchu's Science-Based Industrial Park.
Seven years later, Advanced Wireless Semiconductor Corp. (AWSC) opened its facility in the southern city of Tainan. Today, AWSC is the largest GaAs foundry in Taiwan, producing approximately 1200 100mm wafers/week. The company plans to open a new 150mm foundry in 2003.
In 1999, the world's first, pure-play 150mm GaAs foundry was established in Taiwan by WIN Semiconductors. WIN produced an impressive record of advancement, progressing from ground breaking to first product wafer in a little over a year's time. WIN's factory at Linkou currently produces about 200 150mm wafers/week, but plans are in place for an expansion to 2000 wafers/week (10% of the estimated world market) by 2005.
Global Communications Taiwan (GCT) was founded in 2000 by shareholders of California's Global Communications Semiconductors (GCS). It is currently operating a small, 100mm wafer foundry in Hsinchu, but plans to open a new 150mm facility by year's end. GCT claims that its new facility will have a maximum capacity of 1250 wafers/week.
Compound Semiconductor Manufacturing Corp. (CSMC), based in Hsinchu, was founded in mid-2000. Like many of the other foundries, CSMC is initially focusing on producing HBT power amplifiers for wireless applications. It also plans to manufacture other kinds of optoelectronic devices in the future. While low-volume, 100mm pilot production began earlier this year, predictions call for CSMC to produce 2500 150mm wafers/week by 2004.
 A highly skilled work force makes TSMC (shown here) and other Taiwan chipmakers formidable competitors in world markets. |
Originally called Taiwan Compound Semiconductor Co. (TCSC), Suntek Compound Semiconductor Co. was established in August 2000 by Procomp Informatics and Acer Communications and Multimedia. Although the first 100mm production wafers will not be processed until the end of this year, Suntek has already announced significant partnerships with Mitsubishi and Celeritek. This has led to a projection that Suntek will build a second, 150mm foundry in 2002, with capacity reaching 4500 wafers/week by the end of 2004.
The latest company to enter the foundry market in Taiwan is Transcom Inc. Headquartered in Tainan, Transcom claims to be the island's first supplier of pHEMT technology. Its current 75mm and 100mm manufacturing facility has a capacity of just over 300 wafers/week, which is divided between foundry and production of Transcom's own products. The company hopes to increase the capacity of this facility to more than 700 wafers/week by the end of 2002.
The future of Taiwan foundries
With the slowdown in the GaAs device market, it is likely that the expansion plans of many of these companies will be delayed or may never come to pass. Indeed, with such high levels of competition, it is quite plausible that some of these foundries may not survive the current downturn. The new generation of foundry facilities has only recently begun production and does not yet use large-scale manufacturing methods. Their main concern at present has to be funding, with investment coming largely from GaAs manufacturers in the US and Japan, who are themselves suffering in the recession. In the present financial climate, these "parent" companies need to be sure that there will be a market for these services now and in the future. This will be made all the more difficult by the established foundries, which will almost certainly have learned many lessons about positioning and competition from the current business environment.
The more established foundries may be able to continue operation in spite of the low volumes of wafers currently being run through production lines. Indeed, some of these companies are believed to be pushing forward expansion plans to take advantage of quiet periods when product will not be affected by changes to the factory. Some analysts have gone even further, suggesting that the recession may bring more sources of business to foundries.
Fab operation
Many device manufacturers have seen the output of their fabs reduced dramatically over the last few months. Their process engineers have taken advantage of this downtime and the increased availability of production tools to develop new processing methodologies for manufacturing the next generation of device designs. The fabs hope that these new devices can be developed and manufactured in time to capitalize on the next period of market growth.
Still, there are two large barriers most companies will have to overcome. First, it is estimated that it will take one to two years for the market to return to the levels of productivity realized in 2000. As a result, companies must be prepared to fund the design and development work with little return on capital.
Second, the majority of GaAs IC manufacturers will be selling into the same emerging markets. Therefore, competition for these new devices will be fierce. Chipmakers will have to include additional benefits to the device to ensure that it stands out from its rivals.
Future foundry business?
The costs of operating a fab can be high. With little or reduced production, some fabs may find it is simply more profitable to mothball their own facilities and outsource to foundries, since they would only have to pay for the overhead incurred during IC production. Equipped with the latest process equipment and skilled engineers, foundries are quite experienced in quickly producing working devices from the design stage. This would not appear to be a viable long-term solution, however, because foundry costs will almost certainly increase with the market's demand for product.
Reducing the cost of a product is a good way of making it more competitive. Yet the lack of available funding today makes it almost impossible to invest in new tools or systems that may increase a fab's efficiency. Again, it is the production-oriented foundries that could offer a solution. Many have already invested in efficient production methods in order to capitalize on reduced time to market. During a recession, when competition is at its highest, more value for money paid can become an important device benefit.
Conclusion
Taiwan's GaAs foundries, established for high-volume manufacture of GaAs ICs, are now caught in a climate of low production and minimal investment. Established fabs have shown their operational benefits in a growing market, but their future in the current industry downturn is uncertain. The industry will have to wait and see which companies will emerge from the recession to become the coveted "TSMC of GaAs."
Andy Bavin received his degree in physics from Bath University, UK. He joined Trikon in 1996 as a process engineer in the PVD group, where he worked on the development of Trikon's patented Forcefill and Hi-Fill technologies. He is currently a marketing engineer, responsible for the marketing and sales support of Trikon's PVD equipment. Trikon Technologies Inc., Ringland Way, Newport, UK NP18 2TA; ph 44 1633 474 558, fax 44 1633 414 180, e-mail [email protected].