It s no secret that there s an awful lot of building going on in Taiwan, with new GaAs foundries and epiwafer vendors seemingly springing up all over the place. The exact number of fabs is something of a mystery, and seems to change depending on who you ask, and when you ask. Over the past year, the situation in Taiwan was characterized by investors with large amounts of capital viewing GaAs as an excellent investment opportunity. In some cases, these investors had little or no knowledge of the GaAs industry but were attracted by sky-high market projections and the success of silicon foundries such as TSMC and UMC. With the recent changes in market conditions (either real or perceived), there has been some retrenching by investors, who are being more careful about where they spend their money, with the result that some projects will be delayed or may not get off the ground at all. For the worldwide GaAs industry, this is probably not such a bad thing, since it reduces the threat of an over-capacity situation in the market, which would drive prices through the floor. It also may somewhat reduce the pressure being felt by GaAs substrate suppliers (see page 77). The failure or delay of some of these plans is not surprising when one also considers that there are only a finite number of groups that fully understand complex process technologies like HBTs. Although such processes have matured considerably, it remains relatively easy to get initial results and much more difficult to ramp up to high volume production. Nevertheless, many projects remain on track, aided in no small part by the Taiwanese government, which is aggressively supporting compound semiconductor technologies, according to John Lau of WIT SoundView. "Taiwan s government has targeted compound semiconductors as the next generation technology needed for advanced communication chips," says Lau, who attended the grand opening of the new WIN Semiconductors fab in late December, an event marked by speeches from the Vice President of Taiwan and several ministers of state. "The government has invested indirectly in both WIN and Compound Semiconductor Manufacturing Corp. (CSMC)," he says. "Both these companies are following the highly successful foundry model developed within the silicon industry by companies such as TSMC and UMC." New GaAs Foundries WIN and CSMC join Hexawave and Advanced Wireless Semiconductor Co. (AWSC) on the list of GaAs fabs in Taiwan, and these will soon be joined by Global Communication Technology (GCT), which is eventually planning to build two fabs (see bellow). Two other ventures are moving forward rapidly: G-Advanced Semiconductor Manufacturing Co. (GASMC) held a ground-breaking ceremony for its new 6-inch GaAs fab in July of last year. The first phase of construction at the Koos Lungtan Industrial Park is expected to be completed by April 2001. GASMC, founded by Uni-President Enterprises, Compal and Bao-Chen Group, will start trial production of 6-inch GaAs MMIC wafers in the fourth quarter of 2001. Epiwafer producer Procomp Informatics has set up a joint venture with Acer Communications and Multimedia to build a 6-inch GaAs foundry. One of the foundry s largest customers will be Mitsubishi Electric, while Celeritek has also invested in the new company, known as Suntek Compound Semiconductor Co. Ltd. An interesting point to consider is why these new companies are all offering foundry services. One reason is undoubtedly the optimism generated by the huge success of TSMC and UMC; also, as Ron-Ting Huang of GCT points out, there is a lack of experienced design teams in Taiwan. Another issue is the difficulty faced by new, small companies in trying to gain design wins with large manufacturers of products such as handsetsthe new companies may not have the critical mass to be credible to the major handset suppliers. UMC s chairman, John Hsuan, has expressed reservations about Taiwan s GaAs industry, pointing out that there aren t enough fabless design houses to stimulate process development at the foundries. He believes that if the foundries receive orders only from integrated device manufacturers (IDMs), their business will be "merely a manufacturing businessit would not require continual improvement in technological expertise." Hsuan went so far as to predict that only one GaAs foundry company would survive in Taiwan. Epiwafer Vendors Predictably, along with the new foundry activity in Taiwan there are also a growing number of companies providing GaAs epiwafer services. The list includes Procomp Informatics, VPEC and UEC [see CS 6(7), p.28], as well as new players such as South Epitaxy Co. [see CS 6(7), p.31]. ATMI recently announced that it would expand its Epitaxial Services business into Taiwan and the Asia-Pacific region, although it does not have plans to establish a GaAs epiwafer production facility in Taiwan at present. Among the new epiwafer ventures are the following: The world s largest HBT epiwafer supplier, Kopin, is in the process of setting up an overseas subsidiary in Hsinchu to bring it closer to the local Taiwanese market. Several 6-inch production MOCVD systems are being installed in a leased facility, and production is expected to begin in mid- to late 2001. Hitachi Cable has spent 1.7 billion yen (US$14.5 million) to acquire a 51% stake in Giga Epitaxy Technology Corp., a company which is building a facility to manufacture GaAs-based epiwafers. The 10,000 m2 plant, which is located in Yangmei, Taoyuan County, is expected to become operational in the third quarter of 2001. Giga Epitaxy will initially produce an equivalent of 2,000 4-inch wafers a month, and within five years the capacity is expected to grow to 10,000 wafers/month. The company will focus on the production of 4- and 6-inch HBT epiwafers grown by MOVPE. Hitachi Cable s plant in Japan is currently capable of producing 8,000 wafers a month; the firm plans to triple epiwafer capacity at that plant by June 2001. One of Taiwan s largest silicon foundries, UMC, has provided 40% of the capital to fund Epitech, which plans to make 6-inch InGaP and InP HBT epiwafers. Construction of the company s facility in Tainan began in September 2000, and trial production is expected to begin in February 2001. By 2004, Epitech is targeting a production volume of 150,000 6-inch HBT wafers per year (12,500/month) as well as 24,000 4-inch VCSEL epiwafers. Global Communication Technology Global Communication Technology Corp. (GCT) is close to completing its first 6-inch GaAs foundry in Hsinchu, Taiwan, and is already planning a second fab. Compound Semiconductor spoke to GCT s President, Ron-Ting Huang, a former vice president of Global Communication Semiconductors (GCS). CS. When will the GCT fab be finished? When will you start trial, and volume, production? GCT. GCT s 6-inch GaAs fab, which is equipped with a class 1 clean room, will be completed in January 2001. Equipment installation and qualification will follow right away, and we expect these procedures to be completed in March. Process qualification will be started in Q2. Trial production will commence in Q3 and mass production will begin in Q4, 2001. CS. What will be the capacity of the new fab? GCT. The capacity of Fab I will be 5,000 6-inch wafers per month by the end of 2001. CS. How are plans for GCT s second fab coming along? GCT. Construction of the second fab will begin in Q1, 2002 and is expected to be completed in Q1, 2003. The maximum capacity of Fab II will be 15,000 6-inch wafers per month. CS. What s the relationship between GCS and GCT? GCT. GCT is a completely independent company, and eventually we plan to become publicly traded in either Taiwan or the USA. We expect that GCT and GCS will become a second source for each other. CS. Who are GCT s main shareholders? GCT. First International Computer (FIC), VIA and Central Investment Holdings each own 12.67% of GCT. CS. What are the main reasons to establish GCT in Taiwan? GCT. Manufacturing costs are much lower in Taiwan than in the USA. CS. Why are new foundries appearing in Taiwan, and not integrated device manufacturers (IDMs)? GCT. TSMC and UMC in Taiwan have demonstrated a very successful model for Si foundry services. Hopefully, GCT can demonstrate the same successful model for GaAs. Taiwan still suffers from a lack of good RF design teams, so it makes more sense to start as a foundry. CS. Where will most of GCT s business come from? GCT. We expect a 50:50 split between fabless design houses and IDMs. CS. What factors will make existing IDMs outsource to foundries? GCT. The main factors are cost reduction and flexibility.