Taiwan becomes driving force in single-wafer technology
08/01/2004
Adoption of single-wafer processing technology for semiconductor manufacturing continues to increase worldwide, but the Asia-Pacific region — particularly Taiwan — is outpacing the overall global growth rate.
Taiwan is home to the world's largest silicon foundries, as well as a number of major DRAM manufacturers. Foundries account for 70% of the capacity added in Taiwan since 1995, when single-wafer processing trends began to emerge in new fabs. DRAM manufacturing now represents about 30% of Taiwan's IC manufacturing capacity.
A look at the evolution of Taiwan's semiconductor manufacturing industry can give a good picture of the future of single-wafer wet cleaning. Single-wafer technology's ability to lower costs and deliver production flexibility make it highly attractive to both pure-play chip foundries and DRAM providers for a variety of reasons.
Taiwan's semiconductor manufacturing industry rests on two pillars:
- technology transfer — the historical start of the industry in Taiwan, and currently the basis for DRAM production; and
- the reliance of fabless customers on foundries.
Of these two market segments, foundries serving fabless customers have been the locomotive engine pulling Taiwan's semiconductor manufacturing success.
The fabless business model enables the design originator to focus on design and delivery to customers, while foundries focus on efficiency and economies of scale that enable fabless customers to achieve cost-efficient products. This combination — based on specialization — frees the foundries from nonproduction activities and enables them to build and enhance competitive advantages in manufacturing technologies and production costs.
Taiwan's emergence
As Taiwan's IC manufacturing industry evolved from foreign-owned to locally owned fabs or joint ventures, it grew into a "bipolar" business model: pure-play foundries relying on fabless companies, using the foundry's manufacturing economic advantages and technical flexibility; and memory manufacturers producing DRAMs with licensed technology (mostly from Japan) and selling them to licensors or on the open market.
Taiwan's two leading foundries quickly rose to dominance in the global foundry business, outspending the rest of the foundry industry (particularly over the last five years) to achieve a significant size differential. This growth enabled them to develop the technological infrastructure needed to move quickly and aggressively into more advanced technologies (e.g., 300mm wafers, copper interconnects, and 90nm geometries) and technology nodes ahead of the International Technology Roadmap for Semiconductors (ITRS) (Fig. 1).
 Figure 1. Silicon foundries pull ahead of the IC roadmap. (Source: FSA) |
Single-wafer technology's ability to lower costs and deliver production flexibility makes it highly attractive to both foundries and DRAM providers. Foundries, however, have been particularly critical to Taiwan's rapid conversion to single-wafer processing. The technology's promise of lower costs and higher flexibility naturally attracted foundries, which were early adopters of single-wafer tools. The first such tool was installed in Taiwan in 1996, and as Fig. 2 illustrates, the single-wafer installed base in Taiwan grew at an average of 30%/year from 1996 to 2003.
 Figure 2. Taiwan's single-wafer wet-processing tool installed base, 1996–2003. (Source: SEZ estimates) |
Taiwan's memory makers followed the general trend of DRAM manufacturing, with a delayed adoption of more advanced technology nodes compared to foundries, and with an emphasis on keeping costs down and stalling change for as long as possible. Thus, traditional wet benches were replaced by single-wafer tools in memory fabs more slowly than in logic-oriented foundries. For both foundries and memory providers, the main applications for single-wafer clean tools were backside clean and backside film removal. In recent years, polymer removal also became a key application for these tools.
Current trends
With more than 1200 fabless companies worldwide — more than 480 (40%) of which are in Asia, according to the Fabless Semiconductor Association — Taiwan's foundries have access to both a large world market and a critical mass of regional and even local fabless companies that are both customers and technology accelerators. In fact, it is estimated that Taiwan produces two-thirds of all wafers manufactured for fabless companies.
With their combination of advanced technology, flexibility, and economies of scale, Taiwan's foundries have also attracted outsourcing business from integrated device manufacturers (IDMs). The FSA estimates that about 30% of most pure-play foundries' business comes from such outsourcing. At times, the foundries' reliance on single-wafer wet-clean tools for certain processes has an impact on IDMs' own production and increased use of such tools.
By the end of 2003, Taiwan's total wafer production was estimated to be about 60% from foundries and 40% from memory makers. Taiwan's DRAM makers produce about 45% the output of Korea, the world's number one DRAM producer. While this is a significant market position, it is also a drawback. As part of their efforts to overcome this size disadvantage vis-à-vis Korea's DRAM manufacturing, Taiwan's memory producers have accelerated their adoption of single-wafer processing as they move to smaller geometries. Lower costs — particularly costs of consumables — have been, and remain, the main driver. By the end of 2003, Taiwan's installed base of single-wafer wet-processing tools was the largest in Asia-Pacific, with the highest growth rate in terms of units (Fig. 3).
 Figure 3. 2003 Asia-Pacific single-wafer wet clean tools installed base. (Source: SEZ estimates) |
In recent years, the emergence of foundries and increased IC production in mainland China has raised some questions about the manufacturing markets in Asia. While Chinese foundries try to provide both logic- and memory-making capacities, they appear to be in the formative stage of their development — both technically, reaching a stable and consistent manufacturing technology and performance, and in terms of establishing a sustainable business model and long-term relationships with fabless customers. At the end of 2003, it was unclear what kind of threat China's foundries would pose for Taiwan's memory makers and foundries.
Looking into the future
The FSA estimates that by 2010 about half of all IC revenues will be generated by operations that exploit the fabless business model. That represents an increase of more than 300% in 10 years, up from just 15% in 2000. Such a scenario would ensure the continued survival of Taiwan's pure-play foundries, further encouraging foundry investments in single-wafer wet clean tools, particularly as these manufacturers move to the 65nm technology node. By 2006, the installed base of single-wafer wet clean tools is expected to increase by ~50% (Fig. 4).
 Figure 4. Taiwan single-wafer clean tools expected installed base, 2004–2008. (Source: SEZ estimates, conservative scenario) |
As to Taiwan's memory makers, alliances and continued emphasis on accelerated technology investments are key factors to remaining competitive globally. As both sectors of Taiwan's IC manufacturing enter the 65nm technology node, the importance of single-wafer clean tools is expected to grow significantly, pushing both foundries and DRAM makers into additional investment in such tools. Front-end-of-line (FEOL) single-wafer clean tools are expected to enter the market in the next two years and grow rapidly from there. This development would likely expand the installed base of these tools to an estimated 380 by 2008.
Foundries are expected to continue as early adopters of new or newer technologies, and are thus anticipated to maintain a 70% market share of the single-wafer wet-clean tool installed base until 2008. Memory makers are expected to move more aggressively into single-wafer wet clean tools, starting in about 2006; by 2008, they may represent close to 30% of the installed base in Taiwan.
The number of vendors sharing this market is also expected to increase. Most in the industry agree that single-wafer's importance will escalate rapidly at 90nm, with 65nm marking the inflection point at which single-wafer technology becomes a critical, indispensable factor of manufacturing success. As competition rises, Taiwan's IC manufacturing industry stands to be the main beneficiary.
David Chen received his bachelors and masters from Tam Kang U., and his PhD in solid-state physics from the U. of Texas in Austin. He is VP of technical marketing and account general manager in Taiwan for SEZ Asia-Pacific, 13 F, No. 93, Shui Yuan St., Hsinchu, Taiwan, R.O.C.; ph 886/3-572-7890, fax 886/3-572-4567, e-mail [email protected].
Hai Benron is a graduate of Stanford U., and received his MBA from Yale U. He is VP of business marketing and account general manager in China for SEZ Asia-Pacific, 25D, No. 138 Pudong Ave., Pudong, Shanghai, China 200120; ph 86/21-5882-1908, fax 86/21-5882-2098, e-mail [email protected].