ASIAN CASE RESEARCH JOURNAL, VOL. 16, ISSUE 1, 183–224 (2012)

AcRj
Taiwan DRAM Industry in 2009
This case was prepared by Professor Chien-Nan Chen of National Dong Hwa University, Taiwan and Professor Chengli Tien of National Taiwan Normal University, Taiwan, as a basis for class discussion rather than to illustrate either effective or ineffective handling of an administrative or business situation. Please address all correspondence to Professor Chien Nan Chen, Department of Business Administration and Graduate Institute of Logistics Management, National Dong Hwa University 1, Sec. 2, Da Hsueh Rd., Shoufeng, Hualien 97401, Taiwan. E-mail: cn_chen@ mail.ndhu.edu.tw

DRAM, considered as strategic materials of the electronics industry, is like water and cannot be monopolized by Korea. Global technology companies pay much attention to how the Taiwan government supports its DRAM industry. Taiwan DRAM industry is still likely to be profitable if the Taiwan government considers its investment cautiously to help DRAM companies merge or transform1. — Frank Huang CEO of Powerchip Semiconductor Corporation Dynamic random access memory (DRAM) production and wafer foundries became two major pillars in Taiwan’s semiconductor industry beginning in the 1980s. Mosel Electronics, founded in 1987, and Acer Semiconductor Manufacturing, founded in 1989, were among the early DRAM firms in Taiwan. United Microelectronics Corporation (UMC), founded in 1980, and Taiwan Semiconductor Manufacturing Company (TSMC), founded in 1987, were among the early wafer foundry firms in Taiwan. In 1991 Mosel Vitelic Inc. was established through a merger of Mosel Electronics and Vitelic to develop its DRAM products actively with its establishment of the 6-inch fab in 1993; thus it was the first DRAM manufacturer in Taiwan. During the years 1993–1995, the DRAM market boom and supply shortages encouraged firms in Asia to invest in the DRAM industry. During this period, a number of Taiwanese firms started to team up with leading international
1B.

S. Su, 2008. Frank Huang: the whole world is watching how Taiwan saves its DRAM industry. Commercial Times, December 5. © 2012 by World Scientific Publishing Co. DOI: 10.1142/S0218927512500083

184

ACRJ

DRAM firms and served as original equipment manufacturers in exchange for technology. These alliances include Nanya Technology with OKI (Japan), Winbond Electronics with Toshiba (Japan), Powerchip Technology and Vanguard International Semiconductor with Mitsubishi (Japan), and Mosel Vitelic Inc. and ProMOS with Infineon (Germany). However, excessive investment fueled by these alliances led to recession. During the years 1996–1998, the bust of the semiconductor industry caused the leading DRAM manufacturers from Japan and the United States to exit the DRAM industry or slow down their investment. However, on the contrary, Taiwanese firms increased their investment and successfully filled the spots previously held by these leading firms. For example, Texas Instruments (TI) exited the DRAM industry and Acer became the only owner of Acer Semiconductor Manufacturing, which later was acquired by TSMC in 1999 and became a dedicated foundry. Mosel Vitelic exited the DRAM market in 2003. A number of Japanese firms also exited the DRAM businesses, and only Elpida, merged from NEC and Hitachi, survived. The market bust also affected the Korean firms and Hynix was merged from Hyundai and LG to become one of the only two DRAM manufacturers in Korea (the other one was Samsung Electronics). Sluggish investment during the early 2000s triggered the market boom during the years 2005–2006. Oversupply of the DRAM industry in the end of the year 2008 was the result of the global crisis and supply expansion in the years 2007–2008 (Exhibit 1). The market price of DRAM was lower than cost, which caused manufacturers’ cash outflow and loss (Exhibit 2). Moreover, DRAM manufacturers carried many outstanding corporate bonds and bank loans (Exhibit 3). Therefore, they turned to local governments for financial help due to incapability to pay the debts. Among these requests for financial help in all countries, Taiwan’s bailout plan received a great deal of attention. Taiwan was the second largest DRAM manufacturing country in 2007 (Exhibit 4). The total production of six DRAM manufacturers (Powerchip, Rexchip, Nanya, Inotera, ProMOS, and Winbond) in Taiwan was only behind Korea.

TAIWAN DRAM INDUSTRy IN 2009

185

However, Taiwan DRAM manufacturers did not have their own technology, but depended on foreign technical partners instead. Therefore, this financing assistance from Taiwan’s government was likely to impact the benefits of domestic and international companies. Taiwan’s government hoped to alter its DRAM industry and to make it more competitive by means of this financing. Chi-ming yin, former Minister of the Ministry of Economic Affairs of the Republic of China (ROC), once said that
The government is unable to interfere with any individual DRAM company. We only focus on the whole infrastructure. We have several plans, including consolidating all companies into one. However, this is difficult. Alternatives are to select one company as a base platform, to establish a new company or to divide existing companies into two groups. Each plan has its advantages and disadvantages. We have to find the most efficient and easiest solution to solve the current DRAM difficulties costeffectively2.

On March 6, 2009, the Ministry of Economic Affairs established a company named Taiwan Memory Company (TMC) as the platform for industrial renovation. John Hsuan, vice president of UMC, was selected as the convener, and either Micron or Elpida was to be chosen as TMC’s strategic partner. This was to improve and sustain Taiwan’s DRAM technology by controlling the intellectual property (IP), processing, and technology3. John Hsuan said that
Taiwan needs to have enough capital and experts to compete with Samsung. We have manufacturing advantage. However, we are weak on design and development. We hope to start to work on the design and development in the next two months4.

y. Hou, 2009. Restore the scene: a sense of humor or misunderstanding? Commercial Times, February 14. 3H. H. Lu, 2009. Which side to take: USA or Japan? The merger deal about DRAM was finalized in 3 months to establish Taiwan Memory Company. John Hsuan is appointed to be the convener. Commercial Times, March 6. 4S. Z. Lin, 2009. John Hsuan recreates DRAM; Government only needs to invest 30 billion NT dollars. China Times, March 11.

2y.

186

ACRJ

The setup of TMC, however, has brought about much controversy. First was the gigantic gap between the setup of TMC and what DRAM companies expected from the Taiwan government. These DRAM companies expected that the government could help transfer the technology, not set up another company to compete with. In addition, between Micron and Elpida, which one was TMC going to pick as its strategic partner? How was the whole industry going to be affected by this decision? How would TMC effectively utilize limited government funding? John Hsuan needed to work out these issues in the shortest possible time. Otherwise, as the DRAM price was to rise, the intention for integration from these DRAM companies would diminish. chARAcTeRIsTIcs of sTAnDARD DRAM PRoDucTs AnD ITs InDusTRy DRAM was a kind of memory integrated circuits (IC), which amounted to about 50% of memory market from 2006 to 2010 (Exhibit 5)5. DRAM and central processing units (CPUs) were necessities for computers. DRAM was used as a medium to temporarily store commands and information by computers. Generally speaking, the faster the DRAM, the faster the computing. Under guidelines of Moore’s Law that the quantity of transistors on an IC doubles every 18 months, the DRAM industry continuously pursued faster speed from early Extended Data Out (EDO), to Synchronous DRAM (SDRAM) and to Double Data Rate (DDR) in the year 2000, followed by DDR26. DRAM products were divided into standard products and specialty products. Standard DRAM products were used in computers, such as notebooks, desktop computers, industrial computers, and servers. Specialty products were used as registers in electronic appliances, such as interactive
5Memory is different from storage device. Memory is characterized by faster access speed and limited storage capacity. It is needed to execute programs and commands and data in process will be saved in the memory. Those programs and data saved in the storage device need to be kept in the long run. 6C. C. Wang, 2006. 12-inch wafer fabs become the key market competition in DRAM. CTimes, September 5.

TAIWAN DRAM INDUSTRy IN 2009

187

television, video games, personal digital assistants (PDAs), set-top boxes (STB), broadband modems, and printers. The demand of computer products amounted to 80% of total DRAM production in 2009 (Exhibit 6). The high demand of computer products in the past had brought growth of DRAM (Exhibit 7). The principles and process to manufacture DRAM were similar to general IC products. The circuits were etched or stacked through ion implantation, diffusion, and thin film procedure on blank wafers. This process was usually referred to as the front-end IC manufacturing. Separate dice, cut from wafers during the front-end manufacturing, were then assembled and tested (known as back-end manufacturing) to become the chip for sale (Exhibit 8). A general personal computer in 2008 needed 2GB of standard DRAM, which was assembled by 16 pieces of 1GB DRAM chips. The characteristics of a standard DRAM product were short-term steady supply, large demand, and standardized specification. Thus, manufacturers created cost privileges and entry barriers by expanding scale, using large-sized wafers, introducing advanced processing technology, and increasing yield rate. The larger the wafer size, the more dice were cut from a wafer. Thus, average die cost was reduced. The more advanced the process manufacturing was (narrower interconnect), the more dice a wafer could produce, and average die cost was reduced as well. Similarly, higher yield rate brought a higher quantity of dice and lowered unit cost (Exhibit 9 and Exhibit 10). On the other hand, while smallersized wafer fabs expanded their scale to a larger size, not only were larger wafers produced, but technology also advanced. Therefore, expanding scale increased production, and if yield rate was well maintained, larger-sized wafer fabs owned absolute competitive advantage (Exhibit 11). Technically, the scale went through major changes every two years. From years 1996 to 2001, there were eight generation upgrades from 0.40 μm to 0.15 μm. The scale change brought production capacity increased by 7 times7.
7J. M. Tang, 2005. The DRAM industry’s 10 year rise and decline — an exploration of Japanese business management’s difficulties in coping with the trend of worldwide competition. unpublished master’s thesis, EMBA, National Taiwan University.

188

ACRJ

As far as scale benefit was concerned, compared to 0.11 μm technology, 90 nm was expected to increase output by 58%. 90 nm changing to 70 nm could increase output by 55%. 50 nm produced more than double the output of 70 nm. Cost was dramatically lowered with advanced processing technology. Therefore, once processing was upgraded and yield rate was increased to the level that was more cost effective than the old processing system, the manufacturers started their transformation to new processing. From years 1996 to 2001, scale processing was advanced averagely every half-year. Since 90 nm had entered the stage of high yield rate, the manufacturers aimed their transformation to 50 nm to obtain first-mover advantage (Exhibit 9). Building wafer fabs and purchasing equipment was costly. Under monthly production of 30,000 pieces, 6-inch, 8-inch, and 12-inch needed USD 500 million, USD 1.2 billion, and USD 2.5 billion funding capital respectively8. The cost was highly relevant to the processing technology. The production needed to be in the latest technology to keep ahead, yet expensive equipment raised high risks to a wafer fab, and therefore resulted in high entry barriers. Take 12inch wafer fabs for example. These manufacturers treated depreciation differently: some used average depreciation for 4–8 years, while others used accelerated depreciation for 2–4 years. In general, depreciation took 40-50% of total manufacturing cost (Exhibit 12). In addition to a high investment, a wafer fab required a longer preparation time. It took approximately 18 months for a 12-inch wafer fab to be completed from plant design to production9. Wafer fabs had to plan the production rate precisely or looked for strategic partners to share the risks in order to avoid early production in an immature market or late production in an overly competitive market.

Industry yearbook 2008. Department of Industrial Technology, the Ministry of Economic Affairs. ROC 9M. Chin, 2006. A model for setting up 12-inch wafer fabs from the perspective of photolithography engineering — based on a new fab in Tai-Chung Scientific Park. unpublished master’s thesis, Department of Industrial Management, National Taiwan University of Science and Technology.

8Semiconductor

taiwan dram industry in 2009

189

moreover, standard dram products were highly dependent on computer products and the demand for them was determined by computer product sales. Computer market cycles affected price fluctuations, and thus, DRAM manufacturers’ profit. Due to the high building cost of a wafer fab, a booming market encouraged dram manufacturers to produce more, but recession brought low production. However, due to the lagged effect of investment, the demand-supply gap occurred in the meantime. upgraded processing brought increased production, but if demand failed to follow the increased supply, recession appeared. thus, wafer fabs reduced their investment or production and waited for demand to rise. the cycle therefore repeated. during the years 1991–2007, the dram industry had experienced recession three times (Exhibit 13). The first recession happened during the years 1996–1998. it occurred due to excess supply following dram booming during the years 1992–1995, and substantial increases in production capacity for 6-inch converting to 8-inch wafers. the second recession happened during the years 2000–2001 and resulted from the dot-com bubble. in 2001, dram price increased again, and the following booming four years encouraged more plant expansion, capital spending, and production capacity that caused more supply than demand for dram and led to the third recession.

Global Market CoMpetition intel conducted dram commercial production in the 1970s. us companies including international Business machines (iBm), ti, and motorola dominated the market in the early 1970s, followed by Japanese companies. Japanese semiconductor companies had over 50% market share under toshiba, nEC, Hitachi, mitsubishi, and Fujitsu, and ranked first worldwide. Korean companies rose in the 1990s, and forced major Japanese companies to exit the market or to merge. Until the late 2000s, Samsung still ranked first in the market by a more than 30% market share. taiwanese dram companies, including Powerchip, nanya, and PromOs

190

ACRJ

entered the mainstream market in the mid 1990s (Exhibit 14). Taiwanese companies had a 13% market share for the first two quarters of 2009 (Exhibit 15) while their production capacity of 12-inch ranked first worldwide (Exhibit 16); they had 19 plants on mass production while 11 plants among them were for DRAM10. DRAM had become an oligopoly after long-term evolution. It was dominated by four companies: Samsung, Micron, Elpida, and Hynix. Samsung was the only selfproduction company while the other three were in a strategic partnership with each other to share expensive R&D and construction cost. The four companies are introduced as follows:

samsung electronics Samsung Electronics entered the DRAM industry in 1983 and appealed to a number of top talents with the brand impression of Korea’s Number One. Samsung developed the world’s first 256Mb DRAM in 1996 and had been the world leader in the DRAM industry since then. With design and development skills and high market share, Samsung had made Japanese companies exit the market they had previously dominated, just like American firms exiting the DRAM industry in the 1980s11. Since the 4th quarter of 2007, Samsung had had a hard time with the DRAM business, just like other companies. However, with profit coming from other businesses, Samsung was not prone to decreased production, but invested more capital instead. They lowered their profit margin and increased production of 12-inch wafer from 66% of first quarter in year 2007 to 83% of fourth quarter in year 2008 to accelerate the phase-out of 8-inch. Samsung strategically excluded other competitors by means of the recession12.
10C.

H. Chou, 2008. Live or die about DRAM: Competition between countries. Economic Daily News, November 14. 11Samsung Electronics. Case published by Harvard Business School (9-705-508). 12K. C. Peng, 2008. Strategic alliance in DRAM: Unpredictable game. Economic Daily News, December 22.

taiwan dram industry in 2009

191

Micron Technology Team micron technology was the only dram company in the united states. its ranking of semiconductor patents was only behind intel. micron had the most patents among all dram manufacturers13. in production, micron had nine plants of 8-inch wafer fabs and one plant of 12-inch wafer fab. However, it had high production cost due to outof-date 8-inch wafer fabs, which had become its burden (Exhibit 14). Without sufficient support of 12-inch production capacity, micron had a cost disadvantage. thus, micron had aggressively turned to other strategic partners for 12-inch production. in march 2008, micron and nanya established a strategic partnership to develop technology under 50 nm. initial cooperation was launched at nanya’s 12-inch fab. in the meantime, nanya closed the long-term cooperation with Qimonda14. during the partnership with micron, nanya had acquired 50 patents and free rights to use the rest of micron’s patents. nanya had acquired a total of over 1,100 patents within the united states, Japan, and taiwan, which exceeded Elpida15. in november 2008, micron acquired 35.6% stock share and production capacity of inotera from Qimonda16, which had even tightened the business relationship with nanya17. Elpida Team dram had been in recession since 1996 and the recession had made many Japanese companies exit the market. Hitachi
y. Peng, 2009. technological alliance: micron is better. Commercial Times, march 6. L. Ho and C. H. Chou, 2008. Big shuffle in DRAM: Nanya Technology allies with micron. Economic Daily News, march 4. 15s. H. Lee, 2009. the patents owned by nanya technology outweighs Elpida. Commercial Times, august 4. 16inotera is a 12-inch wafer fab, co-founded by nanya technology of the Formosa Plastics Groups and Qimonda, each owning 35.6% of stock shares. the monthly production of inotera is up to 120000 pieces, which importantly support the productions of nanya technology and Qimonda. 17y. L. Ho, 2008. 400 million us dollars: micron buys 35% of inotera’s stock shares. Economic Daily News, October 14.
14Y. 13H.

192

ACRJ

and NEC spun off their DRAM business divisions in year 1999 and merged them to form Elpida, which later received partial stock share from Mitsubishi. Since then, there had been only one DRAM manufacturer in Japan. Although Elpida only had an 8-inch plant in the beginning, it utilized integration and technology transfer strategies and further acquired investment from Intel and Kingston to move on to specialty memory products. Standard products were delivered from Powerchip, a foundry in Taiwan. Powerchip and Elpida had formed a strategic partnership in technology development, production, and sales. In December 2006, Elpida and Powerchip co-established Rexchip and announced the establishment of four 12-inch plants within 4–5 years with a total investment of USD 15 billion, in which each company shared 50% investment and production capacity18. In November 2008, Elpida acquired Rexchip as a subsidiary by increasing Rexchip’s stock share from 48.8% to 52% by means of supporting Powerchip. Frank Huang, CEO of Powerchip, however, was still the CEO of Rexchip. Rexchip was one of the most advanced DRAM plants in Taiwan. Its monthly production capacity was approximately 80,000 pieces, and it owned the latest 65 nm technology19.

hynix Team Hynix had a similar background to Elpida. Encouraged by the Korean government, it was merged from Hyundai Electronics and LG Semiconductor in 1999 and was renamed Hynix after becoming independent from Hyundai in 2001. Hynix was unable to move to 12-inch technology due to its financial weakness. The international capital market had described Hynix as a company in technically bankrupt status. However, Hynix still existed with substantial help from the

18C.

H. Chou, 2006. A big investment of NTD 450 billion: Taiwan adds 4 more wafer fabs. Economic Daily News , December 8. 19 y. L. Ho and C. H. Chou, 2008. Elpida raises its investment: Rexchip becomes a Japanese factory. Economic Daily News, November 28.

taiwan dram industry in 2009

193

Korean government. its dram products had been the subject of an anti-dump tax imposed by the united states, European union, and Japan due to the acceptance of government aid. Hynix and PromOs signed a strategic partnership contract in 2003. thus, in 2004 when the market rebounded, Hynix received half production capacity from PromOs’ second 12-inch plant, and moved to 12-inch generation in 2005. in may 2008, PromOs announced the signed contract with Hynix that Hynix would transfer 50 nm technology to PromOs and Hynix would invest 8% to 10% in PromOs, and would become PromOs’ second largest stockholder 20. However, in July 2009, PromOs’ cooperative relationship with Hynix was terminated, but PromOs still needed to pay royalties to Hynix. Ben tseng, Vice President of PromOs said:
PromOs signed the technology transfer contract with Hynix in year 2005 with a focus on 50 nm, and the contract is expected to end in year 2013. according to the contract, PromOs takes over Hynix’s technology and reserves 50% production capacity for Hynix. However, for future consideration, we terminate the cooperation relationship with Hynix to release the 50% production capacity that is reserved for Hynix21.

From market share (Exhibit 15), samsung took about 30% and ranked first. Hynix and ProMOS total took approximately 20%. micron, nanya, and inotera total approximately accounted for 20%. Elpida, Powerchip, and rexchip approximately accounted for 20%. the market was quite obviously dominated by these four groups. in addition, taiwanese companies generally were unable to develop technology due to low market share and insufficient resources. taiwan’s primary advantage came from both production capacity and capability. that is, yield rate and efficiency of most Taiwan DRAM companies were beyond their technical partners. therefore, taiwan companies usually

20C.

H. Chou, 2008. PromOs’ partner for 50 nanometer: Hynix is picked. Economic Daily News, may 10. 21s. H. Lee, 2009. Cooperation ends with Hynix: 50% of production capacity of PromOs is resolved. Commercial Times, July 24.

194

aCrJ

acquired the technology from their technical partners in exchange for production capacity. according to the ministry of Economic affairs, during the years 2006–2008, the Formosa Plastics Group’s nanya technology and inotera memories had paid technical cooperation and royalty fees up to ntd 27 billion, while Powerchip technology had paid about ntd 24.4 billion and PromOs had paid about ntd 22 billion. Frank Huang said, the Japanese government has invested much resource and manpower in development. But what Elpida wants is only production capacity. the cooperation that government approves is that Elpida cooperates with rexchip for rexchip’s low cost production, but technology stays in Japan. the Japanese government is cautious in revealing technology. therefore, it is out of the question to acquire technology without paying the royalties22.

Unbalanced Demand and Supply and Impact of the Financial Tsunami in 2008 with the expectation of the microsoft Vista operating system bringing new computer demand, dram companies had been in expansion mode since 2006. However, the market did not go as expected, which caused excess supply over demand. the dram market was then in an awful imbalance of demand and supply. Price of dram 667 mhz 1 GB dropped almost 75%, which was from original usd 2.29 to usd 0.58 (Exhibit 17), which was even below manufacturers’ cost of USD 1. The total loss from the first quarter to the third quarter was over USD 8 billion globally. The financial tsunami that followed in the fourth quarter aggravated the situation for many dram manufacturers in crisis23.

22s.

H. Lee, 2009. the r&d center of tmC is said to go to umC. Commercial Times, april 25. 23H. H. yeh, 2008. the war of life and death in the 2008 dram industry has become a war between countries. cnyes.com, december 30. http:/ /news.chinayes.com/newsbase/20081230/web1302.shtml

taiwan dram industry in 2009

195

total iC industry production in taiwan in year 2008 was twd 654.2 billion, dropping 11.2% compared to 2007. Of this total production, wafer fab production was twd 446.9 billion, slightly dropping 1.1% compared to 2007. Production of self-owned product (mainly in the dram industry) was twd 207.3 billion, drastically declining 27.2% compared to 200724. manufacturers needed to face not only operational loss, but also bonds payable and bank debts with the impact of the recession and over-supply. thus, almost all manufacturers requested short-term financial help, and the competition between manufacturers was elevated to competition between countries.

Qimonda Exited the Market Qimonda was the first manufacturer to exit the market after 2007. Though it received Euro 325 million in financing from the German Saxony Anhalt state government, Infineon, and Portugal Bank, the aid was not sufficient to solve Qimonda’s financial problems and an additional Euro 300 million was needed. The financial aid thus failed. Qimonda filed for bankruptcy with the German government on January 23, 2009. inotera and winbond stopped shipment to Qimonda in the meantime25. their 12-inch plant in Virginia, united states, was shut down and 1,500 employees were laid off. an announcement was made on February 11, 2009 that the production in their 12-inch plant at dresden, Germany, was to be reduced by 75% of its production; that plant ceased its full operation in april the same year. Qimonda’s market share in 2008 was approximately 9.5%, which ranked it one of the major manufacturers in the world. However, misguided technical

24P. y. Lee, 2009. review and outlook of the semiconductor industry during the 4th quarter of 2008. the industrial Economics and Knowledge Center, the industrial technology of research institute. http:/ /www.nettv888.net/ezcatfiles/cust/img/img/52/2009_03_09_001.doc 25Y. J. Chen and C. C. Lee, 2009. Qimonda files for bankruptcy, Taiwan DRAM fears for future. Commercial Times, January 24.

196

ACRJ

development policies26 and insufficient 12-inch fab production capacity had failed Qimonda in the market27.

Powerchip, nanya, and ProMos were settled as full cash Delivery stocks The total loss of Taiwan’s five major DRAM manufacturers was approximately TWD 160 billion and total bank loans were TWD 420 billion, which made up 2.3% of all bank loans in Taiwan. Every manufacturer foresaw the potential financial gap and turned to the government for financial help in the end of 2008. ProMOS was settled to full cash delivery on February 19, 2009, because of the incapability to pay the redeemed Euro-convertible bonds (ECB) payable, USD 335 million scheduled on February 1728. On May 5, Nanya was settled to full cash delivery since stock net value was less than TWD 5. Total loss of Nanya in 2008 was TWD 37.6 billion, followed by a loss in the first quarter of 2009 of TWD 10.5 billion. Nanya had suffered for eight quarters of loss in a row, and its stock net value had become only TWD 3.77. On June 19, 2009, Powerchip was settled to full cash delivery stock due to the incapability to pay ECB for USD 157 million, due on June 17, 2009. Only Inotera and Winbond remained normal in the open stock market29.

to production technologies, DRAM factories can be divided into two trench camp and stack camp. Qimonda represents for the trench camp and the market share of the trench technology camp was once as high as 23% in 2006, but when the technology hits the bottleneck and is behind the technology of the stack camp for about half a year to a year, the market share has declined to 8% in 2008, which means the stack technology reaches the market share of 92%. At the same time, the trench camp also announces the technology of Buried Wordline and formally announces the technology of the trench camp has walked into history. 27P. y. Lee, 2009. Review and outlook of the semiconductor industry during the 1st quarter of 2009. The Industrial Economics and Knowledge Center, The Industrial Technology of Research Institute. http://www.nettv888.net/ezcatfiles/cust/img/img/52/0901001.doc 28T. T. Wang, 2009. ProMOS is classified as full cash delivery; the volume of sale order reaches 200,000 pieces. China Times, February 20. 29H. H. yeh, 2009. The club of full cash delivery club for DRAM: after ProMOS and Nanya Technologies…. Powerchip joins the club on June 19. cnyES.com, June 18. http:/ /news.cnyes.com/dspnewsS.asp?fi=NEWSBASE\20090618\WEB301

26According

TAIWAN DRAM INDUSTRy IN 2009

197

japan Government Paid jPy 30 billion to save elpida As the only DRAM company in Japan and the world’s third largest DRAM company, Elpida faced the same difficult financial situation. The annual statement of 2008, published in March 2009, disclosed the total loss of over JPy 178.8 billion (approximately TWD 60.08 billion)30. On June 30, 2009, the Japanese government decided to offer JPy 30 billion (approximately TWD 10.27 billion) to Elpida. Government and local banks also decided to finance the firm; the total financing value was estimated at over JPy 160 billion (approximately TWD 54.76 billion)31. yukio Sakamoto, CEO of Elpida, said, “In the DRAM industry, if the investment is insufficient, technical development and production capacity will fail the company. The capital will be used in top-end technologies and equipment32.”

samsung, hynix, and Micron were also in loss Compared to other companies that asked for government help, some major DRAM companies, which were also in deficit, tended to make profit from loss sooner due to their large business scale and profit from other product lines (e.g., NAND Flash). Samsung, which had performed well financially, had a deficit in the fourth quarter of 2008. However, after loss for two consecutive quarters, Samsung had profit of TWD 6.3 billion and was the first DRAM company to become profitable from loss. Hynix had shown losses for seven consecutive quarters by the second quarter of 2009. However, with the decrease of excess supply, the loss in the second quarter was down to only KRW 58 billion (approximately TWD 1.527 billion), which was better than the previous quarter with a net loss of KRW 1.178 trillion (approximately TWD 31.3 billion) and also
30S.

H. Lee, 2009. Elpida suffers a big loss of 60 billion NT dollars from last year and this year’s expenses are cut in half. Commercial Times, May 13. 31C. C. Huang, 2009. To save Elpida, the Japanese government throws in 30 billion Japanese yens. China Times, July 1. 32Same as Footnote 34.

198

ACRJ

better than the same month of the previous year with a net loss of KRW 711 billion (approximately TWD 18.7 billion)33. Micron had had losses for ten consecutive quarters by June 2009. But due to production cost that was finally lower than the sale price, the gross profit rate turned positive34. Ever since, Micron had an improved and steady financial status, which was due to sufficient cash on hand, cost reduction, and production reduction.

Ic InDusTRy PolIcy In TAIwAn AnD The DIffIculTIes of TMc Government Policy and Ic Industry Development The Taiwan government launched the semiconductor industry in the 1970s to pursue economic development. In 1976, the Industrial Technology Research Institute’s (ITRI) Electronics Research Center (reformed as Electronics Research Laboratories in 1979) started to launch an IC production plan and signed the contract of Integrated Circuit Transfer Authorization with RCA, bringing in IC 7 μm process technology and setting up an IC demonstration workshop. The Institute actively brought in technology and transferred it to local companies. During the initial period, the government led all projects, invested capital, imported technology, trained people, and took on all risks aiming to construct the infrastructure of the semiconductor industry. After holding the processing technology brought in by RCA with the promotion of VLSI Circuits and Systems Research Project during the years 1983–1988, the ITRI derived private enterprises such as UMC (1980), Taiwan Semiconductor Manufacturing Company (TSMC, 1987) and Taiwan Mask Group (1989). Hsinchu Science Park was established in 1980 to provide facilities, plants, and land to
33S.

H. Lee, 2009. The financial statements of Korean DRAM firms report good results. Commercial Times, July 25. 34C. C. Chang, 2009. For the third quarter, the loss of Micron increases, but gross margin becomes positive: The stock price is down by 3%. cnyES.com, June 26. http:/ /news.cnyes.com/dspnewsS.asp?fi=NEWSBASE\20090626\WEB873

TAIWAN DRAM INDUSTRy IN 2009

199

attract high-tech companies. The government also offered incentives such as tariff and tax deduction to attract private companies moving toward technology-intensive sectors such as upstreaming IC design and downstreaming computer hardware, including personal computers, monitors, terminals, main boards, disk drives, and printers (Exhibit 18). In 1990, the Ministry of Economic Affairs continued to promote the five-year development plan of submicron processing technology by building 8-inch wafer submicron laboratories, developing the technology, training people, and setting up submicron institutions to transfer the technology to private companies in the end. Vanguard International Semiconductor, which was developed from ITRI in 1994, led Taiwan to enter 8-inch generation and encouraged the investment in 8-inch wafer fabs in Taiwan. Private companies such as Nanya, Powerchip, Winbond, ProMOS and Macronix were established to produce DRAM. By 1994, Vanguard had the capability of 0.5 μm and took the lead from other competitors in Taiwan; it was only one level below advanced countries’ 0.35 μm. Ever since, the DRAM industry was promoted by private companies, not by government. ITRI also stopped deriving private companies. These private companies in Taiwan had to start competing with each other for technology and funds. The competition among the private firms heated up. The Taiwanese IC industry became well developed after the mid 1990s and formed a tight network. The government played a supportive and assisting role at this stage with efforts such as the five-year plan of recruiting hightech talent during 1996–2001 to enhance human resources; the mechanism of technological project applications by the Department of Industrial Technology, the Ministry of Economic Affairs to assist private firms’ research and development since 2000; the setup of the Semiconductor Industry Promotion Office in the year 2002 to streamline the coordination of resources from various departments; and the launch of the Semiconductor Institute in 2003 to train and develop talents. The Taiwanese IC industry had been well developed up to the late 2000s with the only professionally built divisions

200

ACRJ

of labor worldwide, and it had performed very well. IC design and DRAM production ranked second globally in 2007. Wafer foundry and IC packaging ranked first globally in 2006 (Exhibit 5 and Exhibit 19). Not only the companies’ own efforts and capabilities, but also the government’s policies and promotion efforts contributed to the success of Taiwan’s IC industry.

The Disputes of TMc The DRAM industry held a strategic position in maintaining the integrity of Taiwan’s PC supply chain. The unbalanced demand-supply and impact of the financial tsunami of 2008 had troubled the DRAM manufacturers whose requests for financial help had received government’s attention. On December 16, 2008 at 11 am, unprecedentedly, Steve Appleton, Chairman of Micron, which is the fourth largest DRAM manufacturer in the world and one of Taiwan’s DRAM companies’ important technical partners, and Mark Durcan, CEO of Micron accompanied by Charles Kau, General Manager of Inotera, visited Chi-ming Yin, former Minister of the Ministry of Economic Affairs, with new 50nm technology transfer document. The meeting lasted for six hours and they did not leave until 5:50 pm. They came to Taiwan for financial aid just like Yukio Sakamoto, CEO of Elpida, who also visited Taiwan a week earlier. Ten minutes after the meeting, at 6:00 pm, Yen-shiang Shih, Vice Minister of the Ministry of Economic Affairs, announced the possibility to finance DRAM companies from the Development Fund with up to TWD 100 billion. — Business Week, December 29, 2008 There were two strategies from the Ministry of Economics Affairs: Urgent funding and integrated innovation. For urgent funding, considered a short-term strategy, the government coordinated with banks to defer loan periods. For integrated innovation, considered a long-term strategy, the government was to encourage the industry to merge to make the industry

TAIWAN DRAM INDUSTRy IN 2009

201

more competitive with government’s technical projects and investment assistance35. On March 5, 2009, the Ministry of Economics Affairs reported to the Executive yuan about the Taiwan DRAM renovation plan. A new company named TMC was to be established with government and private institutions’ investment to reform the DRAM industry of Taiwan. The principles were to establish self-owned technology, to upgrade competi-tive capability and to make sure that government resources were optimized. John Hsuen, Vice President of UMC, was invited as the convener to assist the setup of the management team and to discuss technology cooperation and investment with Micron and Elpida. The Ministry of Economics Affairs emphasized that the DRAM renovation plan was to save the DRAM industry of Taiwan, not any particular individual company36. Since the renovation plan was not meant for any existing DRAM companies to solve their individual financial problems, disputes arose: TMC, which was supposed to consolidate Taiwan’s DRAM companies to help them pass through the recession, ended up on the opposite side — TMC was to break the existing alliance relationship of Taiwan DRAM companies. To compete with TMC, Powerchip even expressed the potential plan to consolidate with Elpida on its own37. John Hsuan said:
TWD 900 billion has been invested in Taiwan DRAM industry with bank loans up to TWD 200–300 billion. Merging now will cause the giant production capacity to become liabilities. TMC is not established to finance but to renovate the industry instead. It is one of the plans
35News

release from the Ministry of Economic Affairs 2008.Information on the DRAM industry by the Ministry of Economic Affairs, ROC, December 16. http:/ /w2kdmz1.moea.gov.tw/user/news/detail-1.asp?kind=&id=16272 36News release from the Ministry of Economic Affairs 2009. Ministry of Economic Affairs is going to establish Taiwan Memory Company (TMC) to reengineer the DRAM industry, ROC, March 5. http:/ /www.moeaidb.gov.tw/external/ctlr?PRO=news.NewsView&id=7890 ° 37S. H. Lee, 2009. TMC will not be engaged in merging, and firms in the DRAM industry are disappointed and rebutted; Frank Huang is angry and will save their own firms by themselves. Commercial Times, March 11.

202

ACRJ

for government to save the DRAM industry, but not the only plan. It is irrelevant to any existing DRAM company. DRAM companies currently face short-term difficulties and have to find a solution on their own38.

Frank Huang doubted:
Government should not ignore the current industry structure and stop its support in the name of industrial renovation. The industry is aware of government’s position and will not expect the cash finance from the government. However, it does not make sense to set up another company to compete with current ones, using citizens’ tax money39.

was elpida looking for other Partners? The primary task to prepare for the establishment of TMC was to ensure the technology source and its partner. After taking the convener job, John Hsuan came to the United States and Japan to visit Micron and Elpida and discussed the possibility of cooperation. Elpida confirmed that it would work with TMC and transfer the technology. This endangered the cooperation between Powerchip and Elpida. Since TMC hoped to acquire Elpida’s technology for free, both sides would cooperate by means of cross-shareholding40. Once Elpida worked with TMC, Rexchip would join the partnership as well. Rexchip was likely to become TMC’s major production since Elpida owned shares of Rexchip, which produced 65 nm processing with a monthly production of 80,000 pieces. On the other hand, though, Micron had shown strong interest in TMC. Due to the contracts with Nanya and Inotera, there could be legal issues if Micron were to cooperate with TMC. On April 9, 2009, Micron authorized
H. Tu, S. H. Lee, and H. y. Peng, 2009. John Hsuan: TMC will not be engaged in merging; firms in DRAM are shocked. Commercial Times, March 11. 39S. H. Lee, 2009. Frank Huang: DRAM industry looking good, and soon by April the DRAM units will be out of stock. Commercial Times, March 26. 40S. H. Lee, 2009. TMC will obtain 10% of Elpida’s stock shares. Commercial Times, April 3.
38C.

TAIWAN DRAM INDUSTRy IN 2009

203

Charles Kau, General Manager of Inotera, and Micron’s representative in Taiwan, Fred Fishburn, to co-announce that Micron would not join TMC41. The future of TMc? On July 21, 2009, the Industrial Development Bureau of the Ministry of Economics Affairs published the latest DRAM renovation plan, which revealed that any company that submitted a renovation plan for government funding needed to file the application in three months. At most two companies were to be awarded. The maximum investment including what was to be invested in TMC was TWD 30 billion while a single company’s limit was TWD 9 billion. The renovation plan terms included: 1) To cooperate with foreign technical mother plants and have technology authorization. 2) To list the next-generation technology to be developed by both Taiwan and foreign mother plants. 3) To list how local talents were trained through developing technology with foreign mother plants. 4) To engage in strategic industry adjustment through conducting acquisitions, merging or reorganizing42. However, considering their existing contracts with other technical partners and financial status, many companies questioned the fairness of the plan and doubted that the plan was in favor of TMC after reviewing the terms of the plan. Zi-jun Du, Chief of the Industrial Development Bureau, said
The government will invest in the company whose DRAM renovation plan fits the government’s expectation. If the company is unable to fit the government’s expectation, but has its own way of thinking, the company can still be on its own43.
41S.

H. Lee and H. y. Peng, 2009. Micro will not join; the DRAM of Taiwan are divided into two camps. Commercial Times, April 10. 42News info from the Industrial Development Burea, Ministry of Economic Affairs, 2009. The reengineering action for the DRAM industry, ROC, July 21. http:/ /www.moeaidb.gov.tw/external/ctlr?PRO=news.NewsView&id=8448 43S. H. Lee, 2009. Reengineering DRAM gives Formosa Plastics Groups a third chance. Commercial Times, July 23.

204

ACRJ

On August 3, 2009, TMC44 was officially registered and submitted a financial assistance request to the Ministry of Economic Affairs. The registered capital to establish TMC was TWD 500,000, and the CEO was John Hsuan with a Board of Directors including Kewen Deng, Taisheng Feng, and Supervisor, Eric Chen45. TMC was to cooperate with memorydesigning companies, such as Elite Semiconductor Memory Technology, Etron, and MemoCom in the initial stage to develop mobile DRAM. TWD 18 billion was to be raised initially with 45% coming from the government and 55% coming from private companies. Stage One was scheduled to complete by the end of 2009 with TWD 11 billion to acquire 9.5% of Elpida stock shares. Stage Two was to raise TWD 7 billion to acquire 10.5% of Elpida stock share. TMC was to hold 20% of Elpida’s stock share upon completion of both stages (Exhibit 20)46. yukio Sakamoto, CEO of Elpida, said,
TMC is a design company with technology, and it has announced to enter the mobile DRAM market. Elpida will fully support TMC. Mobile memory has a strong niche and extensive market application. It has attracted the attention of major memory companies. Elpida will help Taiwan to move on to this industry47.

However, while domestic DRAM companies were still under capital pressure, it was difficult for them to accept that the money TMC raised from government and banks was only to acquire Elpida’s technology48. With the setup of TMC, what strategic choice did the DRAM renovation plan by the Ministry of Economic Affairs offer? How could its Chinese name, Taiwan chiyiti kungssu (Taiwan Memory Company) has been registered, the official Chinese name is renamed Taiwan chuanghsin chiyiti kungssu (Taiwan Innovative Memory Company), but is still called TMC in English. 45T. T. Wang, 2009. Winbond signs cooperative agreements with Qimonda. China Times, August 5. 46y. L. Ho, C. H. Chou, C. F. Tsao, C. C. Fei, M. K. Ho, S. y. Lin, and S. H. Hsia, 2009. TMC raises 18 billion NT dollars; SPIL and Mediatek own the shares. Economic Daily News, August 14. 47y. L. Ho and S. y. Lin 2009. TMC is founded and ally with memory design firms. Economic Daily News, August 5. 48yi. L. Ho and C. H. Chou, 2009. NTD 18 billion dollars to exchange the technology of Elpida , Economic Daily News, August 14.
44Because

TAIWAN DRAM INDUSTRy IN 2009

205

Taiwan manufacturers producing standard DRAM products be affected? Beside technology, what resources did TMC own? Where did TMC’s production capability come from? As a DRAM renovation plan, how feasible was it to set up TMC? The future of TMC was regarded as an indication to scrutinize the success of the DRAM renovation plan by the Ministry of Economic Affairs.

206 ACRJ

Exhibit 1 The Demand-Supply Trend (2007–2009)
1Q07 2Q
1,030 1,159 1,389 1,614 189 777 31 2,026 2,353 2,636 3,022 2,030 2,517 2,721 3,154 0.2 4.7 −26.1 −38.6 9,648 7,555 7,822 6,251 −8.1 −21.7 3.5 −20.1 −3.8 13.7 6,015 6,838 −0.6 −29.9 −18 0 2.9 2.9 2 1.7 1.7 1.5 1.1 7 3.2 4.4 3.1 1.1 2.1 3.3 2 1 −9.1 4,806 1.7 −24 −9.1 3,511 4,060 4,423 4,804 4,806 3,405 4,015 4,334 4,652 4,713 81 28 32 117 233 67 74 70 128 244 876 939 1,050 1,289 1,464 1,578 1,612 1,827 238 281 326 323 374 446 500 473 519 650 795 1,675 1,944 2,243 2,465 2,343 2,552 3,046 3,508 2,969 474

(M pcs-512M Eq, QoQ %, USD) 3Q 4Q 1Q08 2Q 3Q

4QE 1Q09E 2QE 3QE 4QE 2006 2007 2008E 2009E 2010E
5,191 1,034 3,641 175 219 172 8,327 11,449 17,187 1,644 5,944 492 2,437 3,764 8,350 12,680 617 1,222

PC Demand

Graphics Demand

Non-PC Demand

1,991 2,184 2,348 2,035

After Market

DRAM Demand

5,189 6,123 6,827 5,697 10,038 16,406 22,852 34,854 5,221 6,129 6,620 5,576 10,422 16,799 22,776 34,213 0.6 0.9 −7 0.1 1 7.5 −3 1 −2.1 6.2 3.8 3 −3 −10.6 −51.6 2.4 1.5 −49.9 −0.3 1 −34.7 −1.8 0.9 −9.9 4,856 6,129 6,422 33,787 31,275 25,090 22,211 30,071 1 26.2 4.8 26.7 −7.4 −19.8 −11.5 35.4

DRAM Supply

DRAM Supply/Demand (%)

DRAM ASP (512M Eq.)

G/R (%, QoQ/YoY)

−8.5 −27.2 6,952 5,285

DRAM Revenue (M$)

G/R (%, QoQ/YoY)

Data source: KB Investment & Securities, 2008. “The DRAM supply-demand case scenarios for 2009”.

TAIWAN DRAM INDUSTRy IN 2009

207

Exhibit 2 The Taiwan DRAM Companies’ Income Statement (2008/3Q~2009/2Q) (Unit: TWD Million) Companies Quarter Sales Revenue Sales Cost Sales Gross Profit Sales Expense Sales Profit Non-business Revenue Non-business Expense Net Profit Companies Quarters Sales Revenue Sales Cost Sales Gross Profit Sales Expense Sales Profit Non-business Revenue Non-business Expense Net Profit 3Q08 14,970 23,912 -8,942 1,489 -10,431 -1,080 2,220 -15,015 3Q08 11,504 14,767 -3,263 2,383 -5,687 -42 3,040 -8,769 Powerchip 4Q08 5,583 16,332 -10,749 1,159 -11,908 589 12,298 -25,500 1Q09 3,923 7,747 -3,824 1,051 -4,876 606 2,017 2Q09 4,685 12,576 -7,890 1,084 -8,974 690 3,447 3Q08 9,349 12,651 -3,302 1,223 -4,525 -781 3,479 ProMOS 4Q08 4,156 10,335 -6,179 1,184 -7,364 88 6,355 1Q09 1,815 8,774 839 2Q09 2,137 7,598 700

-6,959 -5,461 -7,797 -6,161 1,131 1,929 65 638

-6,286 -11,732 2Q09 8,085 11,853 -3,768 1,435 -5,232 608 1,917 -6,541

-8,785 -13,630 3Q08 10,786 13,287 -2,501 136 -2,637 -807 718 -4,051

-8,596 -2,383 2Q09 7,462

Nanya 4Q08 1Q09 6,134 10,753 -4,620 1,456 -6,037 258 6,107 -11,887 6,172 13,341 -7,169 1,358 -8,347 229 2,395 -10,513

Inotera 4Q08 1Q09 8,253 12,933 -4,681 3,448 -8,128 -107 2,041 6,355

10,837 11,029 -4,482 -3,566 132 132 -4,614 -3,698 154 856 -76 338

-10,279 -5,316 -4,112

Data source: Market Observation Post System.

208

ACRJ

Exhibit 3 The Long-term and Short-term Liabilities of Taiwan DRAM Companies (2008/3Q~4Q) (Unit: TWD Million) Companies Quarter Current Financial Liabilities Short-term Debt Payable Long-term Liabilities Due in One year Sum Long-term Financial Liabilities Bonds Payable Long-term Debt Payable Sum Total Financial Liabilities 8,704 49,053 57,757 8,650 54,756 63,406 0 0 0 67,515 0 42,119 42,119 69,249 20,102 38,000 58,102 69,981 26,280 25,000 51,280 66,228 17,993 33,382 51,375 73,183 19,968 37,032 57,000 76,418 12,547 28,867 41,414 12,786 27,279 40,065 4,500 63,015 67,515 5,212 21,918 27,130 0 11,879 11,879 0 14,948 14,948 3,985 17,823 21,808 3,790 15,628 19,418 Powerchip 4Q08 3Q08 ProMOS 4Q08 3Q08 Nanya 4Q08 3Q08 Inotera 4Q08 3Q08

99,171 103,471

Data source: Market Observation Post System.

TAIWAN DRAM INDUSTRy IN 2009

209

Exhibit 4 The Global Ranking of Taiwan IC Industry and Products (year 2007) Production (US Million) Self-owned IC DRAM Mask ROM Design Manufacture Wafer Foundry Packaging Testing 20,872 7,015 353 12,186 22,460 13,774 7,450 3,350 Market Share 9.6% 22.4% 92.6% 26.5% 10.7% 68.1% 47.6% 67.7% Global Ranking 4 2 1 2 4 1 1 1 Leading Countries USA, Japan, Korea Korea Taiwan USA USA, Japan, Korea Taiwan Taiwan Taiwan

Data source: The Semiconductor Industry Promotion Office, IDB, the Ministry of Economic Affairs.

Exhibit 5 The Global Memory Production (2006–2010)

Data source: “Future development analysis of NAND Flash and NOR Flash.” Chi-liang Mo, Topology Research Institute 2009.

210

ACRJ

Exhibit 6 The DRAM Product Application Market

Data source: “Taiwan’s new global strategy and arrangement path after the DRAM renovation,” Chi-liang Mo, Topology Research Institute 2009.

Exhibit 7 The Global Personal Computer Market (2001–2007) (Unite: Thousand, %)
Quantity Notebook Desktop Server Growth Rate Notebook Desktop Server 5.4 -6.4 0.7 16.7 -1.4 4.0 20.3 5.9 16.5 21.9 7.2 11.4 28.8 8.9 11.3 23.9 6.1 9.4 32.0 3.2 7.3 23.1 2.6 10.5 16.8 2.2 10.7 14.8 1.9 9.5 2001 25,747 105,874 4,760 2002 30,033 2003 37,857 2004 46,138 2005 59,435 2006 73,655 2007 2008(f) 2009(f) 2010(f) 160,423 150,965 11,236

97,192 119,689 139,756

104,436 110,612 118,572 129,119 137,020 141,355 144,961 148,150 4,950 5,767 6,425 7,148 7,820 8,387 9,271 10,259

Data source: The Market Intelligence & Consulting Institute, Institute for Information Industry; Industry & Technology Intelligence Service, the Ministry of Economic Affairs, March, 2008.

TAIWAN DRAM INDUSTRy IN 2009

211

Exhibit 8 The Integrated Circuits Production Procedure

Data source: Introduction and Trend of Semiconductor Industry. (II), Professor Chien-hong Cheng, http:/ /mx.nthu.edu. tw/~chcheng/96class.htm

212

ACRJ

Exhibit 9 The Evolution of Semiconductor Processing Technology

Data source: “10 year rises and declines of the DRAM Industry — an exploration of Japanese business management’s difficulties in coping with the trend of worldwide competition”, Jonq-Ming Tang, MBA, National Taiwan University.

TAIWAN DRAM INDUSTRy IN 2009

213

Exhibit 10 Influences to Die Saw Number by Different Processing Technologies Production (μm) Gross Die Hypothesized yield Rate Good Die (Valid Output) 0.5 100 0.95 95 0.35 210 0.9 189 0.25 400 0.85 340 0.2 630 0.75 473 0.18 800 0.7 560 0.175 1,000 0.7 700

Data source: “Economical analysis of high technology industry,” Shan-chiou Chang, ProMOS, 2006.

Exhibit 11 Cost Difference between 8-inch Wafers and 12-inch Wafers 8-inch wafer Wafer cost Gross Die yield Rate Net Gross Die Die Cost Packaging Cost Testing Cost Other Cost Cost per chip $1,671 541 95% 514 $3.25 $0.213 $0.57 $0.177 $4.29 12-inch wafer $2,547 1297 95% 1233 $2.06 $0.213 $0.57 $0.177 $3.09 -28% -36.4% Difference (%) 52.4% 139.7%

Data source: “Economical analysis of high technology industry,” Shan-chiou Chang, ProMOS, 2006.

214

ACRJ

Exhibit 12 Cost Analysis for the Global DRAM Industry
12-inch/90 nm (DDR2 512 Mb) Bare Wafer Cost (US$) 190–225 12-inch/70 nm (DDR2 512 Mb) 190–225 12-inch/70 nm (DDR2 1 Gb) 190–225 Note Wafer Materials Cost Manufacturing Cost (including equipment depreciation and other non-materials cost) Cuttable Die Quantity yield Rate set at 85% Manufacturing Cost of a good die Assembly Cost Test Cost Total Cost for good dice Total Cost (not including depreciation) Materials Cost (not including depreciation and non-materials cost)

Process Wafer Cost (US$)

1300–1500

1350–1770

1350–1700

Gross Die (Die/Wafer) yield Rate (%) Die Cost (US$/yield die) Package Cost (US$/Package) Final Test Cost (US$/Pact) Part Cost (US$/Part) Total Cost Part Cost (US$/Part) Cash Cost Part Cost (US$/Part) Variable Cost

750–1050 85% 1.4–1.7 0.2–0.28 0.15–0.3 2.05–2.45 1.05–1.2

1350–1420 85% 1.2–1.45 0.2–0.3 0.20–0.35 1.6–1.95 0.9–1.1

650–750 85% 2.2–2.6 0.2–0.3 0.2–0.35 2.7–3.1 1.35–1.6

0.9–1.1

0.75–0.90

1.2–1.5

Data source: Gartner; Industrial Economics & Knowledge Center, Industrial Technology Research Institute (2008/8).

TAIWAN DRAM INDUSTRy IN 2009

215

Exhibit 13 Global DRAM Production (1991–2007) Unit: USD Million

Data source: WSTS, DIGITIMES/Ministry of Economic Affairs (2008/10).

216

ACRJ

Exhibit 14 The DRAM Industry Chronology 1983 1985 1987 1989 Samsung announced 64KB Memory, which surprised the industry. ESI announced DRAM products. Winbond Electronics was established. Taiwan’s government launched a five-year research plan of submicron with investment of $7 billion to develop DRAM. Acer and Texas instruments joint-ventured TI-Acer, was the first DRAM manufacturer in Taiwan. The submicron research plan derived Vanguard International Semiconductor Corporation. Powerchip Semiconductor was established to produce DRAM products in 8-inch wafer. 1995 1996 1999 Nanya Technology and Mosel Vitelic were established. Infineon Technologies was created from Siemens semiconductor section. Powerchip was listed on stock market. ProMOS was established. NEC, Hitachi, and Mitsubishi joint-ventured Elpida. TI-Acer was sold to TSMC due to substantial loss. Korea government dominated the merger of Hyundai Electronics and LG Semiconductor. 2000 2001 Vanguard International Semiconductor announced the transformation to become a wafer foundry manufacturer. Toshiba announced its leaving DRAM market. Winbond announced the production transformation from DRAM to Flash. Nanya Technology started to supply DDR DRAM. Hyundai and LG merged to become Hynix. 2002 ProMOS started mass production in a 12-inch plant, which was the first plant in the world. Micron Technology showed strong interests in acquiring Hynix. Vanguard International Semiconductor announced it would stop producing standard DRAM. Infineon Technologies and Nanya Technology joint-ventured Inotera Memories. 2003 2004 Mosel Vitelic stopped DRAM production due to substantial loss. Vanguard International Semiconductor officially stopped DRAM business.

1994

TAIWAN DRAM INDUSTRy IN 2009

217

2006

Powerchip and Elpida joint-ventured Rexchip Electronics. Infineon Technologies divided the memory section to a new subsidiary named Qimonda with the headquarter in Munich. Winbond and Qimonda announced the processing technology transfer of 80nm and the cooperation protocols of 12-inch plants.

2007 2008

DRAM manufacturers expanded 12-inch plants. Micron agreed to buy 35.6% stock share of Inotera Memories previously owned by Qimonda. Micron and Nanya announced to joint venture 50nm DRAM design company. Hynix acquired 8.59% stock share of ProMOS by participating ProMOS’ private placement. ProMOS requested financial help from Taiwan’s government.

2009

Qimonda announced bankrupt. Samsung’s financial report first showed loss during past nine years. The Ministry of Economic Affairs, ROC, invested TWD 30 billion trying to renovate the DRAM industry.

Data source: Business Next (2009/6).

218

ACRJ

Exhibit 15 The Market Share of the Primary DRAM Companies Ranking 1 2 3 4 5 6 7 8 9 10 Companies Samsung Hynix Elpida Micron Nanya PSC Winbond ProMOS Etron Qimonda Others Total
Data source: DRAMeXchange.

Country Korea Korea Japan USA Taiwan Taiwan Taiwan Taiwan Taiwan Germany

Revenue (USD Million) 2Q09 1,179 928 745 551 230 126 80 58 52 0 94 4,043 1Q09 838 705 496 483 171 91 74 48 35 149 89 3,180 4Q08 999 843 643 689 177 78 67 111 38 330 104 4,080 2Q09

Market share 1Q09 26.4% 22.2% 15.6% 15.2% 5.4% 2.9% 2.3% 1.5% 1.1% 4.7% 2.8% 100.0% 4Q08 24.5% 20.7% 15.8% 16.9% 4.3% 1.9% 1.6% 2.7% 0.9% 8.1% 2.6% 100.0% 29.2% 22.9% 18.4% 13.6% 5.7% 3.1% 2.0% 1.4% 1.3% 0.0% 2.3% 100.0%

TAIWAN DRAM INDUSTRy IN 2009

219

Exhibit 16 The Global Primary DRAM Manufacturers’ Production Capacity (Unit: Thousand Pieces/year) Supplier Samsung Area Kiheung Kiheung Kiheung Kiheung Kiheung Hwasung Hwasung Hwasung Hwasung Hwasung Hwasung Austin Austin Hynix Ichon Ichon ChungJu ChungJu Ichon Ichon Usi Usi Eugin ProMOS Hsinchu Hsinchu Taichung Taichung Nanya Taouan Linkou Linkou Inotera Linkou Linkou Fab10 Fab11 Fab12 Fab13 Fab14 Fab15 SAS-12” SAS M6 M7 M8 M9 M10 M11 Ushi 1 Ushi 2 Fab1 E1 Fab1(Fab3-1) Fab2(Fab3-2) Fab3 Fab4 Fab1 Fab2 Fab3 8 8 8 12 12 8 12 8 8 12 12 12 8 8 12 12 12 Plant Fab6 Fab7 Fab8 Fab9 Wafer 8 8 8 8 12 8 8(12) 12 12 12 12 12 8 2004 780 840 780 765 – 780 900 399 90 – – – 510 852 891 780 264 18 – – – 480 480 198 – – 357 474 – 117 – 2005 780 1,050 1,050 1,020 24 1,035 1,260 630 510 120 – – 615 1,050 1,320 825 759 216 – – – 540 480 240 8 – 360 480 – 555 – 2006 960 1,200 1,200 1,200 210 1,260 1,335 861 762 675 30 – 705 408 1,583 1,382 1,260 542 – 270 30 644 480 240 270 – 360 600 – 720 3 2007 960 1,305 1,305 1,305 330 1,365 510 1,020 1,020 1,095 555 – 780 – 1,281 1,671 1,605 1,215 – 849 864 780 486 240 645 15 360 600 – 720 465 2008E 960 1,320 1,320 1,320 360 1,260 600 1,380 1,350 1,260 990 450 780 – 660 1,440 750 1,725 150 630 1,650 405 195 240 750 165 195 180 270 702 702

220

ACRJ

Exhibit 16 (Continued) Supplier Powerchip Area Hsinchu Hsinchu Hsinchu Hsinchu Taichung Hsinchu Winbond SMIC Hsinchu Tainan Shanghai Beijing Beijing Micron Boise,ID Boise Boise Dominian, VT Manassas Lehi Kumamoto Kumamoto Avezzano S’pore S’pore Qimonda Dresden Dresden Richmond,VA Richmond,VA Elpida Hiroshima Hiroshima Hiroshima Toshiba yokaichi yokaichi yokaichi yokaichi Plant Fab8A Fab12A Fab12B Fab12C Rexchip1 Fab4 Fab5 Fab6 F2 F4 F5 Fab1 Fab2 Fab3 IMFT IMFT IMFT Fab1 Fab2 CIS Fab1 Fab2 Fab1 Fab2 Fab1 Fab2 A-2, 1&2 Fl. E300 A1 E300 A2 y-Cube 1 y-Cube 2 y-Cube 3 y-Cube 4 Wafer 8 12 12 12 12 8 8 12 8 12 12 8 8 8 12 12 12 8 8 8 8 12 8 12 8 12 8 12 12 8 8 12 12 2004 444 378 – – – 240 240 – 186 – – 300 300 600 240 51 – 120 300 504 297 360 447 420 540 – 360 231 – 690 426 – – 2005 432 540 126 – – 240 240 – 240 54 0 300 300 600 240 285 – 120 300 504 360 360 396 456 540 21 360 324 3 930 498 189 – 2006 435 540 501 75 – 240 240 180 240 180 66 300 210 240 30 555 21 120 270 369 600 – 384 480 540 219 219 435 300 960 561 555 – 2007 444 540 540 444 105 195 195 252 180 96 762 300 81 240 – 786 435 120 159 285 540 174 327 480 495 369 75 516 525 960 600 1,380 15 2008E 111 520 527 462 885 – – 165 45 24 155 165 54 135 – 840 720 39 39 90 150 555 45 480 105 480 705 705 705 600 450 1,800 600

Data source: KB Investment & Securities, 2008. “The DRAM supply-demand case scenarios for 2009”.

TAIWAN DRAM INDUSTRy IN 2009

221

Exhibit 17 The DRAM Contract Price Trend

Data source: IBT Securities (2009/7).

222

ACRJ

Exhibit 18 The Hsinchu Science Park’s Industry Classifications (March, 2009) (TWD Million)
Industry Integrated Circuit Computer and Peripherals Communication Opto-Electronics Precision Machinery Biotechnology Sum Others Total Approved % 202 53 45 90 26 29 5 450 45.4 11.9 10.1 20.2 5.8 6.5 194 52 46 85 24 26 427 3 430 Registered % 45.4 12.2 10.8 19.9 5.6 6.1 100.0 1,152,958 124,734 40,676 295,032 11,091 9,232 1,633,722 2,290 1,636,012 Registered Capital % 70.6 7.6 2.5 18.1 0.7 0.6 100.0 773,344 83,500 26,818 234,244 8,169 6,441 1,132,516 1,994 1,134,510 Paid-in Capital % 68.3 7.4 2.4 20.7 0.7 0.6 100.0 77,634 12,467 7,798 27,711 2,141 1,164 128,915 307 129,222 Current Employee % 60.2 9.7 6.0 21.5 1.7 0.9 100.0

445 100.0

Note: Current employees include foreign labors: 3329. The communication industry includes 3 closed companies. Data source: The Science Park Administration.

taiwan dram industry in 2009

223

Exhibit 19 taiwan’s iC industry Position in the Global market

14.2% (16.2%) 8.0% (9.0%) 6.9% (7.6%)

note: Value within parentheses is the production value in 2006. Data source: The Semiconductor Industry Promotion Office, IDB, the Ministry of Economic Affairs.

224

ACRJ

Exhibit 20 The Initial Framework of TMC

Government Fund and Investors DRAM Design Firms ESMT Etron MemoCom

Clients

Design

Fund

DRAM Products

DRAM Products

Assembly & Test Firms

TMC DRAM process development, Product Development, Design, Manufacturing, Production Authorization Brand Marketing, Technology Authorization, Offer Strategic Partners’ OEM Manufacturing, Technology, Authorization

Product Capacity Partners Fund Technology Elpida Micron IP Process Manufacturing Product Know-how Trade mark Talent Rexchi

Capacity

Nan Ya Inotera

ProMOS Winbond

Powerchip

Data source: yi-lin Ho, Qiao-yen Lin (2008), “A Preliminary Glance at the TMC Framework and the Comparison between Micron and Elpida,” Economic Daily News, April 2, 2008.

Copyright of Asian Case Research Journal is the property of World Scientific Publishing Company and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.