Click for next page ( 246


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 245
Semiconductors JEFFREY T. MACHER DAVID C. MOWERY DAVID A. HODGES University of California, Berkeley INTRODUCTION Often called the "crude oil of the information age," semiconductors are the basic building blocks of many electronics industries. Declines in the price/per- formance ratio of semiconductor components have propelled their adoption in an ever-expanding array of applications and have supported the rapid diffusion of products utilizing them. Semiconductors have accelerated the development and productivity of industries as diverse as telecommunications, automobiles, and military systems. Semiconductor technology has increased the variety of prod- ucts offered in industries such as consumer electronics, personal communica- tions, and home appliances. Global production of semiconductor components grew from roughly $19 bil- lion to $137 billion (in 1997 dollars) during 1980-1997, an annual growth rate of more than 12 percent (see Figure 1~.2 Nevertheless, the U.S. semiconductor in- dustry, which had pioneered the commercial development of this technology, iThe research on which this paper is based was supported by the Alfred P. Sloan Foundation. We are indebted to our fellow participants in the Berkeley Competitive Semiconductor Manufacturing Research Program and especially to its director, Professor Rob Leachman, for invaluable data, advice, and support. We also appreciate the assistance of Jerry Karls and Howard Dicken of Integrated Circuit Engineering, Inc., Doug Andrey and Lynn Lehsten of the Semiconductor Industry Associa- tion, Dan Hutcheson of VLSI Research, Inc., and Jodi Shelton and Debra Scoggin of the Fabless Semiconductor Association in providing data for this paper. This paper has benefited greatly from the comments of Melissa Appleyard, Rose Marie Ham, and Bill Spencer. The authors are solely responsible for any errors or omissions. 2Market share data presented in this paper represent the dollar amount of billings as reported by member semiconductor firms to World Semiconductor Trade Statistics (WSTS), Inc. 245

OCR for page 245
246 150 _` o . _ m ~ 100 o ._ ~ 50 $ .F a) U.S. INDUSTRYIN2000 ~ Row 7~ /' O 1 980 1 981 1 982 1 983 1 984 1 985 1 986 1 987 1 988 1 989 1 990 1 991 1 992 1 993 1 994 1 995 1 996 1 997 Year FIGURE 1 Worldwide semiconductor production, 1980-1997. Source: SIA 1997 Annual Databook; ICE Status: A Report on the Integrated Circuit Industry, 1980- 1998. experienced wide swings in its competitive performance, especially relative to that of Japan, during this period. The 1980s opened with an expanding Japanese presence in memory components, products widely viewed as essential "technol- ogy drivers" for advances in semiconductor manufacturing processes. U.S. firms steadily lost market share to Japanese firms in memory components during the 1980s, and Intel, now among the most profitable semiconductor manufacturers in the world, nearly collapsed in the 1984-1985 industry recession. At the end of the decade, the M.I.T. Commission on Industrial Productivity (1989) suggested that: The traditional structure and institutions of the U.S. [semiconductor] industry appear to be inappropriate for meeting the challenge of the much stronger and better-organized Japanese competition.... The technological edge that once en- abled innovative American companies to excel despite their lack of financial and market clout has disappeared, and the Japanese have gained the lead. By 1989, however, this dismal picture had begun to brighten, and the market position and profitability of U.S. firms have since improved, especially relative to that of Japanese firms. Stronger U.S. performance is revealed in gains in global market share that rest in part on improvements in product quality and manufac- turing process yields. Improved performance also reflects the withdrawal by most U.S. firms from the fiercely competitive DRAM segment of the semicon

OCR for page 245
SEMICONDUCTORS 247 ductor industry. During and after the late 1980s, U.S. firms shifted to logic and microcomponent products,3 where foreign competition was less intense and they could pursue new product opportunities, many of which drew on their proximity to developers of computer software and other complementary products. Japanese firms, facing less progressive domestic computer hardware and software indus- tries, were less successful in product innovation. Other U.S. firms, such as the so-called "fabless" semiconductor firms,4 have entered the industry successfully as specialists in innovative device designs. Meanwhile, the Japanese firms that dominated DRAMs now face a domestic recession and entry by South Korean and Taiwanese firms with low costs and high manufacturing productivity. Entry by non-Japanese semiconductor manufacturers also has expanded export markets for U.S. producers of semiconductor equipment. Although the Asian economic crisis that began in 1997 is likely to depress global demand for semiconductors in the near term and erodes the financial performance of U.S. producers, the relative performance of U.S. semiconductor firms remains strong, as their continuing lead- ership in global market share indicates. Much of the "renaissance" of U.S. competitive advantage in semiconductors thus reflects exploitation by U.S. firms of long-standing strengths in product in- novation. Many of the new opportunities that appeared in the late 1980s for such product innovation reflected developments in other industries such as telecom- munications and computers, in which U.S. firms demonstrated renewed innova- tive and competitive vigor. The repositioning of U.S. semiconductor firms was if anything aided by the U.S. industry's fragmented structure, criticized by the MILT. Commission and others (e.g., Florida and Kenney, 1990~. U.S. semicon- ductor firms' exploitation of new opportunities for product innovation built on an unusual industry structure that distinguishes this industry from its Western Euro- pean, South Korean, and Japanese counterparts. The U.S. semiconductor indus- try is dominated by merchant producers5 rather than by subsidiaries of large, diversified electronics firms. A number of federal government initiatives, ranging from trade policy to financial support for university research and R&D consortia, played a role in the industry's revival, but the specific links between such undertakings as SEMA- TECH6 and improved manufacturing performance are difficult to measure. Col 3Microcomponents include microprocessors, microcontrollers, DSP devices, and microperipheral devices. 4Fabless semiconductor firms design new microelectronic products but subcontract out the manu- facture of these products to firms ("foundries") specialized in their fabrication. 5Merchant semiconductor firms sell most of their production on the open market, in contrast to captive semiconductor firms who produce semiconductor devices principally for internal "parent" systems divisions. 6The SEmiconductor MAnufacturing TECHnology (SEMATECH) consortium was created in 1987 to develop semiconductor manufacturing technology, using a combination of industry and federal government funding.

OCR for page 245
248 U.S. INDUSTRYIN2000 laboration between equipment and manufacturing firms contributed to improved manufacturing performance, but the size of this contribution as well as the factors that produced higher levels of collaboration in an industry long known for its fierce interfirm competition remain uncertain. Industry managers are virtually unanimous in emphasizing that the crisis of the 1980s forced them to devote much more attention to improving their development and management of manu- facturing process technology. But we do not know how much of the overall improvements reflect this renewed focus by managers, nor do we understand why poor performance was tolerated for so long. In a complex industry such as semiconductors, no single explanation for im- proved U.S. performance is likely to suffice. All of the factors discussed above have contributed to this industry's revival, and it is futile to attempt to assign weights to individual causes. At the same time, the foundation for this com- petitive revival is fragile. U.S. producers' success in repositioning their product lines and developing innovative products does not guarantee enduring domi- nance. The M.I.T. Commission's grim diagnosis of the "structural crisis" of the U.S. semiconductor industry does contain important insights; and at least some of the negative consequences of the U.S. industry's unusual structure have not been addressed. Many of the large corporations that supported much of the ba- sic research that propelled the semiconductor industry's early growth have re- duced the scope of their in-house basic research, and public funding for long- term R&D is more uncertain in the wake of the Cold War. Without a clearer understanding of the factors that gave rise to it, maintaining interfirm collabora- tion may prove difficult. This paper surveys the competitive performance of the U.S. semiconductor industry since 1980. The following gives a description of the industry's decline and revival, focusing on measures of financial and manufacturing performance. In Section III, we discuss the changes in technology management that contributed to this revival. Section IV discusses the non-technological factors that affected the U.S. and global semiconductor industries. A short summary and concluding comments are presented in Section V. INDUSTRY PERFORMANCE, 1980-1997 Our discussion of industry performance begins with a summary of the devel- opment of the global semiconductor industry, highlighting trends in the market shares of U.S. and non-U.S. semiconductor manufacturers and semiconductor equipment suppliers during the 1980-1997 period. Our market-share data are measured in terms of revenues and therefore confound trends in output quantity and the price per unit of that output. This effect is not entirely undesirable; one of the primary factors behind the resurgence of U.S. manufacturers' market share in semiconductors is precisely the higher average selling prices of their output dur- ing the 1990s. But we also wish to discuss trends in manufacturing performance,

OCR for page 245
SEMICONDUCTORS 249 and therefore present data in subsequent sections on product quality, yield, and productivity that exclude these price effects. Market Share U.S. Dominance Prior to 1985 The first transistors, and subsequently the first integrated circuits (ICs), were developed and manufactured in the United States primarily for U.S. military and space programs. By the mid-1960s, the computer and communication industries surpassed the U.S. military as the predominant markets for semiconductors, and the market for semiconductor components has been dominated by commercial applications ever since (Tilton, 1971; Braun and MacDonald, 1978~. From the invention of the IC in 1959 through 1985, the combined market share of U.S. producers exceeded that of firms from all other nations (see Figure 2~. A combination of unusual circumstances, including abundant venture capi- tal, widespread licensing and cross-licensing of key patents, and the willingness of U.S. military and space agencies to purchase semiconductor devices from rela- tively new firms, produced an industry structure that by the 1960s contrasted with those of the Japanese and Western European semiconductor industries. The lead- ing commercial producers of semiconductors in the U.S. included a number of "merchant" firms that specialized in semiconductor manufacture. Many of these firms were relatively young, having been founded during the 1950s and 1960s with venture capital financing. In contrast, the Japanese and Western European semiconductor industries were and continue to be dominated by subsidiaries of large, diversified firms in the electrical equipment industries. 60% 40% ~ 80% Ct s o o ~o%X ~ ~ ~ ~ ~ ~ ~ ~ ~ 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 Year FIGURE 2 Worldwide IC production market share, 1970-1980. Source: ICE Status: A Report on the Integrated Circuit Industry, 1976-1982. us ~ JAPAN -war - EUROPE

OCR for page 245
250 Japanese Growth and Dominance, 1985-1990 U.S. INDUSTRYIN2000 The dominant position of U.S. firms was challenged by foreign sem~conduc- tor producers in the late 1970s. Japanese firms had been active in the sem~- conductor industry since the 1950s but lagged behind U.S. firms in product and process technology. But in the m~d-1970s, MITI, NTT (at the time, Japan's state- owned telecommunications firm), and Japanese producers of semiconductor de- vices and manufacturing equipment launched several research programs to im- prove the semiconductor manufacturing capabilities of domestic firms. These initiatives included the well-known VLSI Program overseen by MITI and a paral- lel program for its semiconductor suppliers sponsored by NTT. Paradoxically, the VLSI Program sought to improve Japanese semiconductor capabilities in or- der to strengthen the international competitiveness of Japan's computer industry. These technology development programs focused on memory devices, which were important in computer systems and whose relative design simplicity facili- tated the testing of new process technologies. The market outlook for these de- vices appeared to be favorable, a projection that was amply borne out by subse- quent events. Finally, "Moore's Law"7 provided a clear "roadmap" of the path of future developments in DRAM technology, enabling Japanese firms to focus their efforts to "catch up" in semiconductor technology. In 1977, Japanese sem~con- ductor producers gained a foothold in 16K DRAMs; by 1979, Japanese producers accounted for almost 42 percent of global DRAM sales (ICE Status, 1980~. Japanese producers became the dominant suppliers of memory devices in the industry by the m~d-1980s, and U.S. firms' market share in memory products plummeted from 75 percent in 1980 to less than 20 percent in 1990 (see Figure 3~. U.S.-Japanese competition in DRAM production took on the characteristics of a So% .0 c' ~ 60% 0 ~ tL cn 0 a) , 0%] ~ 0 cat ~ ca oo 0 cat ~ ca oo oo oo oo oo a) a) a) a) a) a) a) a) a) a) a) a) a) Year FIGURE 3 Worldwide memory production market share, 1980-1997. Source: ICE Status: A Report on the Integrated Circuit Industry, 1980-1998. ~ US JAPAN EUROPE X Row 7Moore's Law was articulated in 1965 by Dr. Gordon Moore, one of the founders of Intel, who pointed out that the number of transistors integrated on semiconductor devices tends to double every 18 months.

OCR for page 245
SEMICONDUCTORS 75% ._ ~ ~ ~ 50% ~ ct Q ~ cn cn .~ 0' 25% ~ ct ~ 0% . ~I I I I I I I I I I I r r I 0 Cal ~oo 0 Cal oo oo oo oo oo Year FIGURE 4 Worldwide semiconductor capital spending share, 1980-1997. Source: ICE Status: A Report on the Integrated Circuit Industry, 1980-1998. 25 + US JAPAN EUROPE X Row "capacity race" firms in each nation invested aggressively in production capac- ity for next-generation products. Aided by their superior access to internal sources of finance, Japanese semiconductor manufacturers were able to dominate this investment competition. The U.S. share of capital spending in the world semi- conductor industry declined from nearly 60 percent in 1980 to roughly 30 percent in 1990 (see Figure 4~. During 1979-1990, Japanese producers were first to mar- ket and increased their overall market share with each new product generation (see Table 1~. The enormous capital requirements of the investment capacity race, combined with fierce price competition in DRAMs and a U.S. industry re- cession, forced many U.S. merchant firms, with the notable exceptions of Texas Instruments and Micron Technology, out of the DRAM market by 1985. By 1990, Japanese firms accounted for 98 percent of sales of 4-megabit DRAMs, then the most advanced memory product. Reflecting their declining fortunes in memory devices, U.S. merchant semi- conductor producers lost considerable market share during this period (see Figure 5~. From a leading share of almost 62 percent in 1980, U.S. chipmakers lost roughly 25 percent of the global market over the next nine years, declining to a TABLE 1 Maximum Market Share by Device Type Device type Volume production Maximum market share (%) U.S. Japan 1K 1971 95 5 4K 1974 83 17 16K 1977 59 41 64K 1979 29 71 256K 1982 8 92 1M 1985 4 96 4M 1990 2 98 Source: Dataquest, cited by Methe (1991) and Langlois and Steinmueller (1998).

OCR for page 245
252 75% o 0 ~ 50% ~ s ~ Oh 0 ~ ._ oh U.S. INDUSTRYIN2000 0%1 ~ 0 cat ~ ~ oo oo oo oo oo oo Year r I r 0 Cal r I I . . _ . . . Row X EUROPE FIGURE 5 World semiconductor production market share, 1980-1997. Source: SIA 1997 Annual Databook; ICE Status: A Report on the Integrated Circuit Industry, 1980- 1998. low point of 37 percent by 1989. Japanese semiconductor firms by 1989 ac- counted for more than half of global semiconductor revenues. The Japanese semiconductor manufacturing equipment industry also enjoyed rapid growth during the 1980-1990 period (see Figure 6~. Indeed, the trends in Japanese firms' share of overall capital spending and Japanese semiconductor equipment market share parallel one another closely, since many Japanese semi- conductor firms purchased most of their manufacturing equipment from domestic suppliers. Japanese firms held less than 50 percent of the equipment market in Japan in 1980, but their share increased to 84 percent by 1991 and remains near 75 percent in 1997 (VLSI Research, 1998~. Japanese semiconductor equipment manufacturers increased their global market share from less than 20 percent in 1980 to almost 50 percent in 1990, largely at the expense of U.S. equipment firms, whose market share declined from roughly 75 percent to less than 45 per- cent during the same period (VLSI Research, 1998~. The rapid growth of Japa Europe >< Row 0% . , ~7 ~... . ~or o Cal ~(D 0D 0D oo oo oo oo a) a) a) a) a) Year O Cal ~(D a) a) a) a) a) a) a) a) FIGURE 6 Worldwide semiconductor equipment production market share, 1980-1997. Source:VLSI Research Semiconductor Equipment Consumption and Production by Region, 1998.

OCR for page 245
SEMICONDUCTORS 253 nose equipment firms appears to be attributable to the growth in investment spend- ing by their major customers, rather than MITI initiatives such as the VLSI Pro- gram (Langlois and Steinmueller, 1998~. Equally important, however, was the superior performance and reliability of Japanese equipment. U.S. Revival, 1989-1997 Japanese firms' advances in DRAMs produced widespread concern within the U.S. semiconductor industry and among government policymakers. This dire competitive situation nevertheless began to change in the late 1980s. U.S. pro- ducers reversed their global market share decline in 1990 for the first time since 1975 (ICE Status, 1976-1991~. But this reversal in market share took place in areas other than memory products, where U.S. firms' global market share has grown only slightly since 1990 (see Figure 3~. Much of the improvement in market share resulted from the efforts of U.S. firms to shift their product mix away from low-margin products such as DRAMs in favor of products that enabled them to exploit their strengths in product inno- vation. Having largely exited the DRAM market by 1985, U.S. semiconductor manufacturers in the l990s focused on logic devices and "mixed-signal" and other digital signal processor (DSP) components for the burgeoning market in com- puter networking equipment. Strong demand for these "design-intensive" com- ponents propelled U.S. chipmakers to market share leadership in the global semi- conductor industry by 1993 (see Figure 5~. By 1997, U.S. producers controlled over 50 percent of the global semiconductor market, well above the 29 percent held by Japanese firms. Contradicting the predictions of analysts who argued that DRAM production was an indispensable "technology driver" for semiconductor manufacturing, U.S. firms' enduring market share losses in DRAMs did not pre- vent this revival in their competitive fortunes. New Competition in DRAMS, 1992-1997 The post-l990 decline in Japanese firms' global market share reflected the revival of U.S. firms in new, more profitable product lines, as well as entry by South Korean and Taiwanese firms into the DRAM market. South Korean firms began DRAM production in 1984, and Taiwanese firms had entered large-scale merchant production of DRAMs by 1994. Rather than shifting to logic products, Japanese firms remained in the DRAM business and sought to be technology leaders in introducing next-generation DRAM devices. But a global recession in the early l990s and the subsequent prolonged domestic recession in Japan depressed demand for next-generation memory products. The weakness of the Japanese counterparts of the U.S. indus- tries (e.g., computer networking, Internet applications, and packaged software) that sparked innovation in the U.S. industry also contributed to Japan's misfor

OCR for page 245
254 75% l ~ c ~ ~ 50% 0 s ~ En id: o: 25% o%d ~ = 1989 1990 1991 1992 ~T T 1993 1994 1995 1996 1997 Year FIGURE 7 Worldwide DRAM production market share, 1989-1997. Source: ICE Status: A Report on the Integrated Circuit Industry, 1980-1998. U.S. INDUSTRYIN2000 JAPAN X Row + EUROPE tunes. The situation was made worse by the appreciation of the yen, which placed Japanese firms' memory chips at a competitive disadvantage vis-a-vis those of other DRAM producers in foreign markets. Japanese firms' market share in DRAM products has declined from roughly 70 percent in 1990 to less than 50 per- cent in 1997 (see Figure 7~. Moreover, more intense price competition has reduced the profitability of DRAMs. Their loss of market share therefore understates the financial damage to Japanese semiconductor firms from their focus on DRAMs. DRAMs now are essentially commodity products, and Japan, Taiwan, and South Korea are engaged in a global battle for market share based on low produc- tion costs and high yields. Japan no longer dominates the memory market as it did in the 1980s, having lost market share to Korean and Taiwanese semiconduc- tor firms. The Korean semiconductor firm Samsung now holds the largest share of the global SRAM and DRAM markets, and Korean semiconductor firms oc- cupy three of the top six spots in DRAM sales (see Table 2~. TABLE 2 Worldwide DRAM Merchant Market Sales (Million Dollars) Company Country 1995 1996 1997 Samsung Korea 6,462 4,805 3,550 NEC Japan 4,740 3,175 2,510 Micron U.S. 2,485 1,575 2,003 Hitachi Japan 4,439 2,805 1,950 Toshiba Japan 3,725 2,235 1,750 Hyundai Korea 3,500 2,300 1,650 LO Semicon Korea 3,005 2,005 1,580 Mitsubishi Japan 2,215 1,400 1,150 Texas Instruments U.S. 3,200 1,600 1,100 Fujitsu Japan 2,065 1,350 1,050 Others 4~999 1~880 1~505 TOTAL 40,835 25,130 19,798 Source: ICE Status: A Report on the Integrated Circuit Industry, 1996-1998.

OCR for page 245
SEMICONDUCTORS 255 The rapid growth of Japanese firms' market share during the 1980s relied in part on their reputation for high-quality products. Similarly, the revival of U.S. firms' market share in the late 1980s and l990s rested in part on improvements in the quality of their products. Although the data on product quality are reasonably reliable, the causes of these trends are less easily discerned. We discuss both the trends and the available evidence on factors that lay behind them in the following section. The Quality Challenge From Japan Product Quality In the early 1970s, Japanese firms recognized that improved quality in their semiconductor products could aid entry into the U.S. and global markets. These chipmakers targeted global firms such as IBM and Hewlett Packard, who needed high-quality components for their advanced electronic systems products. Draw- ing in many cases on practices they had long followed in their other manufactur- ing businesses, Japanese semiconductor manufacturers incorporated statistical process control (SPC), total quality management (TQM), and total preventive maintenance (TPM) into their semiconductor operations.8 By the mid-1970s, Japanese firms were applying SPC methods to semicon- ductor processes in fabrication and assembly in order to reduce process variance and defects quality control practices that U.S. semiconductor firms did not pur- sue until well into the 1980s. Japanese semiconductor firms implemented TQM concepts through extensive training of line operators and selective automation of manufacturing to improve process control, material handling, and data process- ing and feedback. Japanese firms also improved the reliability of their semicon- ductor equipment through preventive maintenance and strengthened their rela- tionships with systems-level customers, semiconductor equipment manufacturers, and materials vendors. These internal management practices produced significant quality differences between Japanese and U.S. semiconductor products. Users of U.S. and Japanese devices discovered Japanese memory products had defect rates that were one-half to one-third those of comparable U.S. memory products (Barron,1980~. In 1980, leading Japanese memory producers averaged 160 defect parts per million (PPM) while U.S. semiconductor firms averaged 780 PPM for the same devices (Finan, 1993~. Their skills in managing the development and introduction of new process technologies also enabled Japanese semiconductor manufacturers to "ramp" out- put of new products more rapidly than their U.S. counterparts. Faster achieve- ment of high production volumes gave Japanese firms advantages in defining See Finan (1993) for a more extensive discussion.

OCR for page 245
276 U.S. INDUSTRYIN2000 deed, we noted earlier that U.S. firms faced significant handicaps in the DRAM "capacity races" that developed in the early 1980s. On the other hand, capital expenditures by U.S. merchant semiconductor producers amounted to more than $14 billion in 1996 and $13 billion in 1997 (ICE Status, 1998~. Investments of this size suggest that constraints on the supply of capital to U.S. firms are scarcely binding. In addition, the recent competitive performance of the large firms active in both the Japanese and South Korean semiconductor industnes, especially those specializing in DRAMs, suggests that one can have too much of a good thing. Low-cost capital has been associated with ovennvestment in manufacturing ca- pacity for commodity products that yield low profits. Histoncally, the U.S. semiconductor industry has faced an abundant supply of venture capital (VC). VC funds have supported the foundation of literally hundreds of semiconductor firms since this industry's inception four decades ago. The data in Figure 15 suggest that this high "birthrate," which has contributed significantly to the industry's technological dynamism, shows few signs of de- clining. Especially interesting is the sharp upsurge in new-firm formation during 1983-1985, a period of severe industry recession. Excluding this penod, an aver- age of eight semiconductor startups appear annually (Figure 15~; the majority (70 percent) of these are fabless firms. The VC community has continued to support new fabless semiconductor endeavors, but has been less generous toward sem~- conductor ventures that include manufactunng. Although there are few reliable estimates of the nsk-adjusted cost of capital in the U.S., Japanese, South Korean, and other semiconductor industnes, U.S. firms may well face a higher cost of capital. Nevertheless, any such differential has not deterred the foundation of new U.S. firms, nor has it deterred large-scale capital investments by U.S. firms that have developed successful competitive strategies that rely on their strengths in product design and innovation. For estab- lishedU.S. semiconductor firms, competitive success appears to lead to abundant 25 20 ~ 15 E z 10 5 o 1980-82 1983-85 1986-88 Year FIGURE 15 U.S. semiconductor start-ups, 1980-1994. Source: FSA Fabless Forum (1995) V.2, n.1. 1 989-91 1 992-94

OCR for page 245
SEMICONDUCTORS 277 capital for investment in plant and equipment rather than vice versa. A higher cost of capital may contribute to the low level of investment in long-term, basic research by many U.S. semiconductor firms. Nevertheless, given the competitive realities of this industry, especially the short product cycles and high costs of R&D for maintaining near-term competitiveness, the risk-adjusted cost of capital would have to be very low indeed to produce higher levels of such investment. Trade Policy Among the several government initiatives emerging from the semiconductor industry's turmoil of the 1980s was a sector-specific trade agreement, the effects of which continue to be debated. The Semiconductor Trade Agreement of 1986 responded to accusations by U.S. firms that Japanese DRAM producers were "dumping" their products in the U.S. market.27 By preventing the imposition of heavy antidumping duties on U.S. imports of DRAMs, the agreement sought to avoid a policy that would drive up the domestic prices of components that were essential to U.S. manufacturers of electronic systems, creating strong incentives for them to shift production to foreign locations. A system of "fair market value" prices for DRAMs was created under the terms of the agreement that was in- tended to prevent dumping in the U.S. and third-country markets. The agreement also included an "understanding" that foreign-sourced components would achieve a 20 percent share of the Japanese domestic market within five years. An exten- sion of the agreement in 1991 retained the market share language but dropped the price-monitoring system. Although the agreement was negotiated in response to the competitive crisis facing U.S. producers of DRAMs, its effects on these firms' activities in DRAM production were limited. Most of the major U.S. DRAM producers had exited from this product line by 1985, well before the agreement was finalized.28 The agreement' s price floors and the associated implementation by MITI of controls on production and capacity investment by Japanese DRAM producers, however, had several interesting effects, few of which directly benefited U.S. semiconduc- tor manufacturers or were foreseen in 1986.29 Higher prices for DRAMs pro 27Flamm (1996) provides the most objective account of the agreement, and these paragraphs draw on his analysis. 28The agreement's "price floor" nevertheless may have aided the remaining U.S. domestic producer of DRAMs, Micron Corporation. 29The period following the agreement was also associated with severe shortages of 256K DRAMs, then a vital component of personal computer and other electronic systems. U.S. computer producers, among others, blamed the agreement and the informal, MITI-guided domestic production cartel that oversaw the agreement's implementation within Japan for the shortages. Concern over DRAM short- ages and the alleged Japanese cartelization of the DRAM market (a condition to which U.S. policy, in the form of the bilateral trade agreement, arguably had contributed) led to the proposal by a group of U.S. computer manufacturers to jointly fund the creation of a DRAM manufacturing consortium, U.S. Memories. As supplies of DRAMs became more abundant, this proposal was abandoned in early 1990.

OCR for page 245
278 U.S. INDUSTRYIN2000 vided an opportunity for South Korean firms to expand their production of these devices, sowing the seeds for more intense competition in this product line in the future. Flamm (1996) argues that similar restrictions on production of electroni- cally programmable memory chips (EPROMs) reduced Japanese exports of these devices and enabled U.S. producers of EPROMs to remain in this product line. Although the Semiconductor Trade Agreement may have provided some ben- efits to U.S. EPROM producers, the effects of its pricing provisions seem to have had little effect on the overall U.S. industry. These provisions did not attract U.S. manufacturers back into DRAM production and imposed heavy short-term costs on major U.S. consumers of DRAMs. The market-share provisions of the 1986 and 1991 agreements, however, were eventually followed by a significant in- crease in U.S. semiconductor manufacturers' market share in Japan, and the agreement is viewed as a key factor in expanded Japanese imports of foreign components. In 1992, the foreign share of Japan's domestic consumption of semiconductor components increased beyond 20 percent, and recent data suggest that this share now is at roughly 25 percent (SIA Annual Databook, 1997~. Ac- cording to Flamm (1996), this increase cannot be attributed solely to growth in Japanese consumption of devices (such as microprocessors) in which U.S. firms have a strong competitive advantage but includes significant growth in other prod- uct areas. In other words, U.S. producers increased their Japanese market share in products where they historically had been relatively weak. The agreement's market-share provisions thus contributed to the revival of U.S. semiconductor firms after 1990, but the timing of this revival is such that the lack of such an import target would not have prevented the U.S. industry's recovery, which was well under way by 1990. Antitrust Policy U.S. antitrust policy played an important role in the earliest years of the semiconductor industry, as Bell Laboratories' liberal licensing of the original transistor and related patents was motivated in part by concern over the outcome of the federal government's antitrust suit against the firm that was settled in 1956. The 1956 settlement also led AT&T to manufacture semiconductor devices solely for internal consumption rather than entering the commercial market. These early actions by the technological pioneer in semiconductors powerfully influenced the subsequent development of the U.S. semiconductor industry. The competitive crises of the semiconductor and other U.S. industries con- tributed to a far-reaching shift in U.S. antitrust statutes and enforcement policy in the 1980s. U.S. antitrust policy was widely criticized in the late 1970s for dis- couraging R&D collaboration. The U.S. Justice Department issued guidelines in 1980 that were intended to clarify the antitrust statutes and the Department's enforcement philosophy toward R&D collaboration, in order to remove impedi- ments to such collaborative undertakings. Nevertheless, continuing industry and

OCR for page 245
SEMICONDUCTORS 279 Congressional dissatisfaction resulted in the 1984 passage of the National Coop- erative Research Act (NCRA). The NCRA has been credited with facilitating the formation of SEMATECH, among other industry-wide collaborations, and R&D collaboration appears to have aided the revival of the U.S. semiconductor indus- try. The act was amended in 1993 to extend its coverage to joint production ventures. An evaluation of the "real" effects of the NCRA and the broader shift in antitrust enforcement policy on the U.S. semiconductor industry's decline and revival is difficult without some clearer specification of the counterfactual situa- tion. Would SEMATECH have been formed without the NCRA? Has R&D collaboration contributed to increased market power and/or poorer industry per- formance? Given the size of the firms that joined together to create SEMATECH and the sustained acquaintance of several of them with the federal antitrust au- thorities, the legislative endorsement of R&D collaboration under the terms of the NCRA almost certainly did aid in the creation of this consortium. The semi- conductor industry's performance suggests that R&D collaboration need not re- sult in cartelization and a weakening of competitive forces, although the large share of the U.S. semiconductor equipment market represented by SEMATECH member firms means that this consortium's vertical relationships deserve contin- ued monitoring. Indeed, collaboration may provide one mechanism for combin- ing the benefits of the U.S. industry's atomized structure and technological dyna- mism with those flowing from closer user-supplier relationships. Nevertheless, very few production joint ventures have been formed since the passage of the 1993 amendments to the NCRA, suggesting that this policy shift thus far has had little effect. Intellectual Property Rights Since 1980, the U.S. semiconductor industry has experienced considerable change in another important aspect of the public policy environment, intellectual property rights. Shifts in U.S. policy toward intellectual property rights began with the 1982 legislation that established the Court of Appeals for the Federal Circuit (CAFC), which strengthened the protection granted to patent holders.30 The U.S. government also pursued stronger international protection for intellec- tual property rights in the Uruguay Round trade negotiations and in bilateral ven- ues. These shifts in federal policy toward intellectual property rights involved both stronger international and domestic enforcement and a somewhat more fa- vorable attitude in the judiciary and antitrust enforcement agencies toward patent 30According to Katz and Ordover (1990), at least 14 Congressional bills passed during the 1980s focused on strengthening domestic and international protection for intellectual property rights. The Court of Appeals for the Federal Circuit created in 1982 has upheld patent rights in roughly 80 percent of the cases argued before it, a considerable increase from the pre-1982 rate of 30 percent for the Federal bench.

OCR for page 245
280 U.S. INDUSTRYIN2000 holder rights. The shift was particularly significant for the U.S. semiconductor industry, because of the relatively limited economic role historically occupied by formal instruments of intellectual property protection such as patents. As we noted in our summary of the industry's development, the unusual circumstances of the industry's founding years, especially the extensive cross-licensing by Bell Labs and the Defense Department' s requirements for "second-sourcing" of many devices, meant that knowledge flowed relatively freely among firms in the indus- try. Interfirm knowledge flows were further enhanced by high levels of labor mobility, by reverse engineering by firms of one another's chip designs, and by diffusion of advances in process technology through equipment suppliers. In addition to these shifts in federal policy affecting all U.S. industries, the semiconductor industry was the beneficiary of a law designed to strengthen pro- tection of industry-specific intellectual property. The Semiconductor Chip Pro- tection Act (SCPA) of 1984 established protection for the design or "mask work" used in semiconductor manufacturing (Stern, 1986~.3i Passage of the S CPA es- tablished a new form of semiconductor intellectual property a "chip design right," best described as a sui generis mode of protection that combines elements of patent and copyright principles with elements of trade secret law (Brown, 1990~. The SCPA extends protection from copying to the three-dimensional images or patterns formed on or in the layers of the semiconductor component that is, the "topography" of the chip and provides for a reverse engineering clause whereby a competitor may reproduce a mask work for the purpose of analyzing it.32 Although it is an interesting experiment in sui generis protection of new forms of intellectual property, the SCPA's economic significance appears to be limited. Only one case has ever been litigated under its provisions.33 The SCPA's unanticipated insignificance appears to be one result of the increasing complexity of manufacturing process technologies in the semiconductor industry. Copies of a device design and mask work are necessary but by no means sufficient to enable large-scale production of infringing products (Kasch, 1993~. As a result, semi- conductor firms during the 1980s and 1990s continue to rely on trade secrets and patents, the value of which has increased as a result of the policy shifts noted above.34 3iMask works represent the three-dimensional pattern of the layers (the topography) of a semicon- ductor component. 32Mask works may be reproduced for the purpose of teaching, analyzing, or evaluating the concepts or techniques embodied in the circuitry, logic flow, or organization of components. Legitimate re- verse engineering may incorporate the results without infringement into another mask work to be produced and distributed. 33This case, Brooktree Corporation v. Advanced Micro Devices, resulted in the award of $26 mil- lion in damages for AMD's infringement under the SCPA and several patents. 34The registration of mask works under the SCPA provisions has advantages over patent filings, which require the disclosure of proprietary information and a time-consuming search through prior art to assert validity. Mask work filing provides immediate registration at minimal cost without a time- consuming search.

OCR for page 245
SEMICONDUCTORS 281 The SCPA nevertheless may play an important, albeit unintended, economic role in the fabless segment of the U.S. semiconductor industry. In order to main- tain short design cycles, fabless firms must extensively reuse design data, "port- ing" designs from one product to another, or contracting with another firm for all or some part of the design. Reusable design components are generally referred to as "intellectual property (IP) blocks," and protection for these IF blocks under the S CPA facilitates the licensing process. The growth of licensing of IF blocks has supported further specialization by some design firms in specific components of overall device designs. These "virtual companies" operate by licensing their pro- prietary designs and architecture to other semiconductor design firms that pro- duce an integrated design, contract with a foundry, and, in many cases, market the final product. The broader shift of federal policy toward stronger enforcement of patent holder rights has been associated with a dramatic increase in patenting and licens- ing among integrated semiconductor manufacturers in the U.S. industry.35 Li- censing has become an important component of profits for some leading manu- facturers. The royalty income of Texas Instruments has grown from roughly $200 million in 1987 to more than $600 million in 1995 (Grindley and Teece, 1997~. Other firms, such as Intel, IBM and AT&T, now rely on licensing to generate revenues and protect product and process technologies. The historic strengths of U.S. firms in product design and rapid innovation should be reinforced by stronger enforcement of patents and trade secrets. The distribution of these benefits within the industry, however, is less clear. Stronger intellectual property protection appears to have benefited established firms. Intel's strong position in its microprocessor product line relies in large part on the firm's intellectual property rights. Another historic strength of the U.S. industry, however, is the ease with which new firms can enter. The effects of stronger intellectual property rights on rates of new-firm formation and entry are less clear. On the one hand, new firms with strong patent positions often find it much easier to attract financing. On the other hand, the costs, in terms of litigation and patent prosecution expenses, of establishing such a patent position are very high. The empirical evidence on the social benefits from stronger intellectual property pro- tection is thin and equivocal. Certainly, the increased litigiousness of established U.S. semiconductor firms has attracted criticism from other U.S. semiconductor producers. In the semiconductor industry, as in others, the U.S. is conducting an experiment in the effects of stronger intellectual property protection, and the im- plications of these new policies for long-term industry performance are surpris- ingly uncertain. 35The number of patents granted in the category "Semiconductor Devices and Manufacture" in- creased from 1655 in 1981 to 5427 in 1994 (U.S. Department of Commerce: Patent & Trademark Office, 1995).

OCR for page 245
282 U.S. INDUSTRYIN2000 CONCLUSION Forecasts of the impending demise of the U.S. semiconductor industry in the late 1980s were considerably overstated. After declining through much of the 1980s, U.S. semiconductor firms undertook corrective actions on several fronts. They exited from product lines in which their historic skills at product innovation provided limited competitive advantage and their foreign competitors' superior access to capital made long-term competition difficult. U.S. firms also improved their product quality and appear to have enhanced their manufacturing perfor- mance, narrowing the gaps between them and foreign competitors, rather than moving ahead. The results of these steps have been dramatic. The U.S. semicon- ductor industry has regained its formerly dominant global market share, and the financial performance of U.S. semiconductor manufacturers now outstrips that of their South Korean and Japanese competitors. Moreover, the revival of the U.S. semiconductor manufacturing industry has reinvigorated the U.S. semiconductor equipment industry. Simultaneously, the South Korean and Japanese firms that specialize in the production of DRAMs are experiencing serious financial losses. In many respects, the revival of the U.S. semiconductor industry relied on the elements of its structure that were the target of criticism in the 1989 report of the M.I.T. Commission on Industrial Productivity. The structure of the U.S. semiconductor manufacturing industry remains very different from that of the Western European or Japanese industries, although the structure of the emergent Taiwanese semiconductor industry is based on the U.S. model and still bears a passing resemblance to it. Populated by numerous, comparatively small, highly innovative firms, and exposed to competition by new entrants pursuing new prod- uct opportunities and new approaches to the semiconductor business, the U.S. industry remains adept at product innovation and rapid strategic repositioning. In addition, U.S. firms have relied on collaboration among semiconductor manufac- turers, and between manufacturing firms and suppliers of equipment, to improve their manufacturing performance. The links between the collaborative initiatives of the 1980s and 1990s and the industry's improved performance remain elusive, however, and further research on these issues is essential if the current strengths of U.S. manufacturers and equipment producers are to be maintained. Although the M.I.T. Commission's overall prognosis of the industry's future prospects was inaccurate, its analysis of the U.S. industry's weaknesses in manu- facturing and long-term R&D investment highlighted other issues that could lead to future competitive difficulties. The very best U.S. semiconductor manufactur- ers appear to be capable of matching the yield and productivity of the best non- U.S. producers, but there is little evidence of consistently superior U.S. manufac- turing performance. As a result, U.S. firms are likely to do best in periods of rapid innovation, especially because of their ability to exploit their presence in one of the world's most dynamic markets for applications of new products that use semiconductor components. But U.S. firms may have trouble competing on

OCR for page 245
SEMICONDUCTORS 283 the basis of their manufacturing skills alone and therefore are likely to face chal- lenges in future periods where they and foreign competitors are pursuing incre- mental innovations within a well-defined technological "trajectory." The U.S. industry is enormously effective in exploiting scientific advances for rapid com- mercialization but may underinvest in the basic research supporting these ad- vances. This is a serious issue for debate, although the recent performance of the much larger Western European and Japanese firms in this industry that have made such investments suggests that simply creating large, diversified firms is an inef- fective solution to this problem. From its very inception, the U.S. semiconductor industry has had close rela- tionships with federal government agencies in charge of R&D and procurement programs. Like other post-war U.S. high-technology industries, the U.S. semi- conductor industry benefited from large-scale investments in defense-related R&D in both industry and academia as well as the procurement programs of federal military and space programs in the 1950s and 1960s. Federal policies in other areas, such as antitrust and trade policy, also have affected this industry throughout its history. But during the 1980s, apart from steel and automobiles, the semiconductor industry was almost without peer in the attention devoted to its welfare and competitive prospects by federal policymakers. The record and legacy of federal intervention in this industry during the 1980s has been criticized by many observers. Nevertheless, one of the most remarkable features of federal policy in semiconductors was the rejection of some alterna- tives that almost certainly would have been far worse for the industry's competi- tive prospects. For example, consider the costs and consequences of a public- private venture like U.S. Memories, specializing in DRAMs, during the l990s. Policymakers and industry managers might well have faced some very unpleas- ant choices between erecting trade barriers against competing imports or allow- ing this venture to slide into insolvency. The proposal of the National Advisory Commission on Semiconductors for a government-backed Consumer Electronics Capital Corporation, which would have been charged with financing the revival of a U.S. industry to consume the products of the domestic semiconductor indus- try, experienced an even more rapid and fortuitous demise. In hindsight, the avoidance by federal policymakers in the Executive and Congressional branches of government of programs that would involve the support with public funds of specific designs of commercial products was wise and consistent with well-estab- lished principles of technology policy. The revival of the U.S. semiconductor industry is an impressive feat, for which government policymakers and industry managers, engineers, and research- ers should share in the credit. But the unexpected nature of this revival, its rather complex causes, the contributions to it of cyclical factors, and the fragility of its foundation all suggest that competitive strength in this industry cannot be taken for granted. Indeed, some foreign producers, notably Taiwanese semiconductor firms, now are entering markets traditionally dominated by U.S. producers, a

OCR for page 245
284 U.S. INDUSTRYIN2000 development that will intensify pressure on U.S. firms and increase the impor- tance of manufacturing performance for competitive leadership. In other words, U.S. semiconductor firms must maintain their strategic agility and strength in product innovation while avoiding significant erosion in their manufacturing ca- pabilities in order to maintain their strength. This task will require imagination and collaboration among government, industry, and academia. REFERENCES Appleyard, M.M., N. Hatch, and D.C. Mowery. "Managing New Process Introduction in the Sem conductor Industry," forthcoming in Corporate Capabilities and Competitiveness, G. Dosi, R. Nelson, and S. Winter, eds. London: Pinter. Barron, C.A. (1980). "Microelectronics Survey: All That Is Electronic Does Not Glitter," Economist 1 March. Braun, E., and S. MacDonald. (1978). Revolution in Miniature: The History and Impact of Semicon ductor Electronics, Cambridge: Cambridge University Press. Brown, H. (1990). "Fear and Loathing of the Paper Trail: Originality in Products of Reverse Engi- neering Under the Semiconductor Chip Protection Act as Analogized to the Fair Use of Nonfic- tion Literary Works," Syracuse Law Review. Burrows, P. (1992). "Bill Spencer Struggles to reform SEMATECH," Electronic Business May 18. Cole, R.C. Managing Quality Fads: How American Business Learned to Play the Quality Game, New York: Oxford University Press, forthcoming. Erickson, K., and A. Kanagal. (1992). "Partnering for Total Quality," Quality, Sept. Fabless Semiconductor Association. (1997). "State of the Fabless Business Model," mimeo, Septem ber. Finan, W. (1993). "Matching Japan in Quality: How the Leading U.S. Semiconductor Firms Caught Up With the Best in Japan," M.I.T.-Japan Working Paper. Flamm, K. (1996). Mismanaged Trade? Strategic Policy and the Semiconductor Industry. Washing- ton, DC: Brookings Institution. Florida, R., and M. Kenney. (1990). "Silicon Valley and Route 128 Won't Save Us," California Management Review 33(1). Grindley, P., D.C. Mowery, and B. Silverman. (1994). "SEMATECH and Collaborative Research: Lessons in the Design of High-Technology Consortia," Journal of Policy Analysis and Manage ment. Grindley, P., and D.J. Teece. (1997). "Managing Intellectual Capital: Licensing and Cross-Licensing in Semiconductors and Electronics," California Management Review. Hatch, N.W., and D.C. Mowery. (1998). "Process Innovation and Learning by Doing in Semiconduc- tor Manufacturing," Management Science, forthcoming. Integrated Circuit Engineering Corporation (ICE). (1987, 1988, 1990-1998). "ASIC Outlook: An Application Specific IC Report and Directory," Integrated Circuit Engineering. Integrated Circuit Engineering Corporation (ICE). (1997). "Cost Effective IC Manufacturing 1998- 1999," Integrated Circuit Engineering. Integrated Circuit Engineering Corporation (ICE). (1997). "Memory 1997," Integrated Circuit Engi- neer~ng. Integrated Circuit Engineering Corporation (ICE). (1998). "Memory 1998," Integrated Circuit Engi- neer~ng. Integrated Circuit Engineering Corporation (ICE). (1997). "Microprocessor Outlook 1997," Integrated Circuit Engineering. Integrated Circuit Engineering Corporation (ICE). (1998). "Microprocessor Outlook 1998," Integrated Circuit Engineering.

OCR for page 245
SEMICONDUCTORS 285 Integrated Circuit Engineering Corporation (ICE). (1976-1998). "Status: A Report on the Integrated Circuit Industry," Integrated Circuit Engineering. Kasch, S. (1993). "The Semiconductor Chip Protection Act: Past, Present and Future," High Technol- ogy Law Journal. Katz, M.L., and J.A. Ordover. (1990). "R&D Cooperation and Competition," Brookings Papers on Economic Activity, Washington, DC: The Brookings Institution. Langlois, R., and W.E. Steinmueller. (1998). "The Evolution of Competitive Advantage in the Global Semiconductor Industry: 1947-1996" in The Sources of Industrial Leadership, D.C. Mowery and R.R. Nelson, eds. New York: Cambridge University Press. Leachman, R., and C. Leachman. (1997). "National Performance in Semiconductor Manufacturing," University of California, Berkeley Competitive Semiconductor Research Program working pa- per (CSM-40). McLoughlin, G.J. (1992). "SEMATECH: Issues in Evaluation and Assessment," Congressional Re- search Service. Science Policy Research Division, #92-749SPR. Washington, DC. M.I.T. Commission on Industrial Productivity. (1989). Working Papers of the Commission on Indus- trial Productivity, Cambridge: M.I.T. Press, two volumes. Methe, D.T. (1991). Technological Competition in Global Industries: Marketing and Planning Strat- egies for American Industry, Westport, Conn.: Quorum Books. National Science Foundation. (1996). National Patterns of R&D Resources, Washington, DC: Na- tional Science Foundation. Rosenbloom, R.S., and W.J. Spencer (1996). "The Transformation of Industrial Research," Issues in Science and Technology 12(3):68-74. Semiconductor Business News. (1998). "Foundries may Build 40 percent of World's Chips by 2010." Semiconductor Industry Association. (1997). "1997 Annual Databook". Semiconductor Industry Association. (1992). "SIA Quarterly Quality Survey." Semiconductor Industry Association. (1997). "The National Technology Roadmap for Semiconduc- tors: Technology Needs," SEMATECH, Inc. Stern, R. (1986). Semiconductor Chip Protection, New York: Law & Business. Takahashi, D. (1998). "Chip Makers Enter Slump; Sales Fall 13 percent," Wall Street Journal July 6. Tilton, J. (1971). International Diffusion of Technology. The Case of Semiconductors, Brookings Institution: Washington, DC. U.S. Congressional Budget Office. (1990). "SEMATECH's Efforts to Strengthen the U.S. Semicon- ductor Industry," Washington, DC. U.S. Department of Commerce: Patent & Trademark Office. (1995). "Technology Profile Report: Semiconductor Devices and Manufacture: 1/1969-12/1994," February. U.S. General Accounting Office. (1992). "Federal Research: SEMATECH's Technological Progress and Proposed R&D Program," July. VLSI Research. (1998). "Semiconductor Equipment Consumption and Production by Region," m~meo.

OCR for page 245