Innovative Flanders: Innovation Policies for the 21st Century: Report of a Symposium (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Session III————————————————————— Cooperative Research and Global Competition in Semiconductors Moderator: Peter Spyns Department of Economy, Science, and Innovation The Flemish Government Current Trends: A U.S. Industry Perspective George Scalise Semiconductor Industry Association Mr. Scalise gave an upbeat assessment of semiconductor market trends, calling it “a great market today,” after 3 solid years of growth. For the current year, he said, the market was forecast to grow by nearly 10 percent, having already grown 8 percent in 7 months. The market was being driven most powerfully by consumer demand for products such as cell phones, digital cameras, digital TV, personal computers, and MP3 players—which accounted for more than 50 percent of demand worldwide. The industrial sector share had dropped slightly below 50 percent. The data going forward, he said, suggested compounded IT revenue growth of about 10 percent “as long as the world economy continues to do well,” especially India and China. Of product areas, MOS logic was by far the most important. Flash memory was replacing rotating memory, he said, a trend that would accelerate, and analog devices were being “pulled along.” Optoelectronics, with growing use in sensors, was becoming a major contributor, while the fastest growing segment was digital signal processing. 64 

SESSION III: SEMICONDUCTORS 65 New Product Trends New products would continue to be cheaper and more powerful, continuing a long trend. Comparing the personal computer of last year with a PC in 1995, he cited a storage capacity 100 times higher and an overall price decline of 98 per- cent. “The overall functionality/cost equation makes it incredibly cheap to buy a PC now,” he said, “and I don’t see any reason why that won’t continue.” Semiconductor technology had entered the nanometer range already, bringing a “whole host of challenges.” In about a decade, he said, the continual shrinking of semiconductors would bring the industry up against physical barriers—power dissipation limits, technological limits, and economic limits—that “may slow us down a little bit.” He cited heat dissipation as a particular problem. But he predicted that the industry was on the right track with both design solutions and process technologies to continues its progress. What Lies Beyond CMOS? Another particular challenge is to find the next generation switch beyond CMOS,12 which he said would probably be required in 10 or 15 years. None of a half-dozen current alternatives to the CMOS logic switch are close to being useful alternatives. With regard to end uses for semiconductors, said Mr. Scalise, product rota- tions were being driven by the consumer now, and product cycles were quickening in response to consumer demand. Cell phone cycles, for example, had dropped in the last few years from 28 to 16 months. Prices had come down, functions had risen, and that trend would continue. To stay in business, companies had to be closely tuned in to what consumers want, and to be in the best position to meet that demand at the right time. As the size of transistors continued to shrink, he said, the industry will have “multi-dimensional innovation requirements.” Today, they are still using equivalent scaling to follow the pace of Moore’s Law. That is, there continue to be new materials and device structures, but still within the existing CMOS scal- ing environment. Something fundamentally new will be required as the transistor passes below 32nm13 and power dissipation issues become acute. Many experts, he said, think that at least part of the problem can be overcome with atomic layer deposition techniques. 12CMOS, or complementary metal-oxide-semiconductor, is the dominant technology mode for digital integrated circuits. The CMOS transistor was invented in 1963. 13The nanometer term describes the size of the smallest feature that can be manufactured on a single chip. There are about three to six atoms in a nanometer, depending on the type of atom. Reducing the size of the features enables smaller, more energy efficient and powerful chips. 

66 INNOVATIVE FLANDERS A “Big Picture” of Semiconductor Research He offered a “big picture” of what the industry is doing to address such chal- lenges. SIA divides its initiatives into competitive (1-3 years), pre-competitive (3-8 years), long-term (8-14 years), and exploratory (15+ years) R&D programs. SEMATECH plays a central role in developing tools and infrastructure, pri- marily in the pre-competitive stage, and the Advanced Transistor Development ­ acility (ATDF)14 makes its fabrication capabilities available to SEMATECH F and others. In addition to SEMATECH, the industry benefits from the Semiconductor Research Corporation (SRC), a group of chip makers and about 100 universities. The SRC’s Focus Center Research Program (FCRP) addresses the industry’s “most intractable problems,” such as the physical limits of silicon, increas- ing product complexity, shrinking design cycles, reduced long-range research budgets, and the dwindling supply of qualified engineers. Its research program involves 5 centers, about 35 universities, 200 faculty, and 400 graduate students to “drive the technology forward and bring out new young talent.” Another SRC program is the Nanoelectronics Research Initiative. Its research objective is to explore ideas and demonstrate proof of concept for a new logic device by 2020. Industry participants include AMD, Freescale, IBM, Intel, Micron, and Texas Instruments. “Structurally,” he said, “we have all the compo- nents covered.” Mr. Scalise concluded by emphasizing that nanotechnology innovation requires the partnership of government (deep expertise in fundamental research), industry (knowledge of technology transfer, road mapping, and the path to com- mercialization), and academia (“out-of-the-box” thinking and new ideas). The NRI now has only about $8 million in annual funding, but during the next 3 to 5 years will scale up to $200 million or so. “We’ll need that to meet the chal- lenges at the nano level.”15 Discussion Dr. Wessner asked whether the federal government was prepared to make the large expected investment in nanotechnology, and whether the effort would be national or international. Mr. Scalise said that the NSF understands the need, but that the next stage, which would require passage of the kind of legislation 14ATDF is an independent subsidiary of SEMATECH’s R&D wafer fab and associated analytical laboratories. According to SEMATECH President and CEO Mike Polcari, “While the SEMATECH consortium continues to focus on our core business of building industry infrastructure in lithography, materials, and manufacturing, the new company represents a complementary effort to meet the more targeted R&D needs of individual companies and universities.” 15A large coalition is also needed to pay for semiconductor R&D costs, which are increasing almost twice as fast as revenues, according to ATDF. 

SESSION III: SEMICONDUCTORS 67 that launched SEMATECH, had not been worked out. He referred to the three pillars of the President’s American Competitiveness Initiative (ACI), and said that the basic research would be ready “when the time comes.” The other two chal- lenges of the ACI are larger. One is to ensure a skilled workforce by that talented students continue to come to the United States and stay here to work along with improving K-12 math and science education. The second is to choose to compete for investment in design and manufacturing projects—the focus of competitors around the world who use incentives and changes in tax policy. China’s Innovation Policies Alan Wm. Wolff Dewey Ballantine LLP Mr. Wolff, who said his work on behalf of the semiconductor industry had taken him to China for the past 10 years, opened with a picture of a billboard near the entrance to the city of Suzhou. On the billboard was written: “Develop- ment is an immutable truth.” That priority has been fulfilled, and has involved very heavy technology-based development. That commitment was described by Jiang Zemin, then General Secretary of the Communist Party of the China ­Central Committee, who said in 1999: “In today’s world, the core of each country’s competitive strength is intellectual innovation, technological innovation, and high-tech industrialization.” This philosophy is pervasive. In contrast with Western leaders’ brief com- ments about innovation in statements of their priorities, said Mr. Wolff, it is a theme in every Chinese leader’s talks. The objective of the strategy is to progress from imitation to production to creating indigenous technology products: “. . . to move from ‘Made in China’ to ‘Made by China.’ It is an objective second to none.” A National Policy of Investment in Technology Among the most significant basic documents describing Chinese thinking is the 15-year “Medium and Long-Term Program on Science and Tech­nology Devel- opment (2006-2020).” This program specifies intensive investments in ­ crucial high-technology products, using policy tools to reward technologies made at home. China plans to increase R&D spending to 2.5 percent of GDP by 2010, which will about equal that of the United States. The State Council of the PRC has issued long lists of technology-based objectives, from core electronic components and new drugs to manned space flight and lunar exploration. The country has pub- lished “guiding opinions” (99 altogether) to move China in a technological direc- tion, advising on such topics as corporate bonds, startup investment funds, debt financing, development zones, and venture capital. According to the 11th Five Year 

68 INNOVATIVE FLANDERS Plan, pieces of which were just emerging, “. . . [China] will promote development by relying on enhancing independent innovation capability, as a national strategy shift in economic growth from relying on the input of capital materials to relying on scientific and technological advancement and human resources.”16 The national IPR strategy, said Mr. Wolff, was to use measures to improve national competitiveness—even if they push against global standards. In the words of one official: “. . . [we shall] abide by international principles and meet the lowest standards of the WTO. . . .” and “[we shall] not only encourage self-innovation, but also encourage absorption, consumption, and innovation of introduced technologies.”17 Measures to Encourage Technology Transfer China has also adopted powerful tools to encourage technology transfer and encourage foreign investment in R&D. One is to exempt from sales tax income earned from the transfer of technology developed exclusively through foreign direct investment in R&D. Another is to give foreign R&D investors with rising development expenses a 50 percent discount in corporate income tax.18 A third is to design procurement regulations that favor domestic products.19 China’s import policy is similarly designed to help China by “watching what comes in and absorbing it.” One ministry recommends increasing “the investment in assimilation and absorption” of imported technologies to “gradually establish a market-oriented system of” technology imports and innovation.20 Other important policies, he said, attempt to guide development. A key one is an antimonopoly policy that aims to “prevent vicious competition in the indus- tries, which if used in a discriminating fashion, could impair foreign investment which has been central to China’s drive to innovate.” In addition, to promote investment, the government provides relief from “social responsibilities.” Local government authorities have set aside billions of dollars to build semiconductor fabs for Chinese companies. Incubation parks are important to the national innovation strategy, and they had, as of 2005, according to Chinese government sources, attracted some thou- sands of companies. The Tianjin Binhai New Area for biotechnology is twice 16Ma Kai, Minister, National Development and Reform Commission, 2006. 17Lu Wei, Deputy Director General, Technical Economic Department, Development Research Center of the State Council, 2005. Foreign R&D investors with development expenses at least 10 per- cent greater than previous year expenses are entitled to a 50 percent discount in total technological development expenses in the current year corporate income tax. 18Guogong Long, “China’s Policies on FDI: Does Foreign Direct Investment Promote Develop- ment?” May 2005. 19Outline of the National Medium- and Long-term Program on Scientific and Technological Devel- opment, 2006-2020, State Council of the Peoples’ Republic of China, 2006. 20Ministry of Commerce, 2006. 

SESSION III: SEMICONDUCTORS 69 the size of the Flemish province of Brabant, said Mr. Wolff, with investments by 69 companies of the Global 500. Shanghai Zhangjiang, a 16-square-kilometer high-tech park, is viewed as China’s “Silicon Valley”—and “Pharmaceutical Valley” as well. Its stated goal is “to form a perfect high-tech innovation chain.” It now hosts 42 foreign companies, including Roche, GlaxoSmithKline, and Medtronic, as well as 70 fabless companies and three semiconductor foundries. China, unlike Japan, has encouraged foreign direct investment as a key com- ponent of its innovation policy. It has also begun to improve IP protection, and has created incentives for indigenous patenting. The number of patents granted had risen 88 percent from 2001 to 2005, and “high-tech” exports have grown rapidly. Conditions that May Hold Innovation Back At the same time, some conditions hold innovation back. Among them, he said, are state planning, the active participation of the Communist Party, and a considerable level of corruption. In some universities, the engineering curriculum requires several hours a week of Marxist philosophy, a distraction from China’s economic goals. Other drawbacks of China’s top-down system are instances of “techno-nationalism,” he said, including attempts at forced technology transfer, favoring domestic companies through national standards requirements, managed trade, misguided industrial policy, and misallocation of capital. While the output of the educational system is massive, some observers have also expressed doubts about the quality of the S&T workforce. According to studies by Duke University, McKinsey, and Cao and Simon, China’s educational system is outdated in emphasizing depth over breadth, a quantitative over a qualitative focus, and neglecting to nurture creativity. Mr. Wolff said that these features tend to produce graduates who do not meet the hiring needs of major western companies. To the country’s credit, however, its leaders are well aware of some of the defects of their system, said Mr. Wolff. For example, the Ministry of Science and Technology lists conditions that hold back the development of sound IPR protection: • No recent history of private property • No history of a culture of IPR • A share of world patents that is still very low • Patent quality that is low • Almost complete lack of patent ownership by Chinese firms IPR Abuses Mr. Wolff mentioned the issue of IPR abuses. According to the State Intel- lectual Property Office, he said, the reason the state does not crack down on 

70 INNOVATIVE FLANDERS counterfeiting is that people would not be able to afford the resulting prices for products. Other deficiencies in IPR protection may be that certain features jeopardize the stated goal of foreign direct investment, including a tendency by Chinese partners to withhold core technologies, limit technology transfer to the routine, hold back key IPR components, and achieve only limited synergies with other Chinese companies. In summary, Mr. Wolff said that China’s innovation system is a work in progress that continues to depend on external input. It is probably held back by the dominant role of the state, he suggested, concluding with a question: Can a government intervene in the market as deeply as China’s does and have a market economy that maximizes innovation? Discussion Dr. Spencer asked how China determines when an invention is Chinese, as opposed to European or American. Mr. Wolff said local innovation is still the exception, but the steady inflow of repatriated engineers from around the world is likely to raise the level of local innovation. Introduction to IMEC Anton de Proft IMEC Dr. de Proft, chairman of IMEC, began by commenting about its success, which he attributed at least partly to a policy that is “kind of hands-on but from a distance.” The government asks for a 5-year program plan, which is followed by in-depth evaluation and the adoption of performance indicators, and then another 5-year program. This, he said, is intended to avoid micro-management. The Goal of Being a Worldwide Center The goal of IMEC, he said, was to be a “worldwide center of excellence” that focuses on exploratory work with a significant impact on industry. Since its founding in 1984, its staff had grown from 70 to 1500, and at the time of the workshop it had about 500 corporate partners. It is subsidized by the Flanders government, contributing 17.8 percent of the budget in 2005. About 22 percent of those working at IMEC are industrial residents who live in Leuven for a year or more, and 14 percent are from academia in Flanders. More than 50 countries are represented, including France, the Netherlands, Germany, Japan, Korea, China, the United States, and many others. IMEC’s basic technology platform, he said, is nanotechnology and its overall mission is “making things smaller, better performing and allowing to address 

SESSION III: SEMICONDUCTORS 71 1984 25 0 2006 Budget: 235 M€ 20 0 Established by state government Millions of Euros of Flanders in Belgium about 280 MUS$ 15 0 Non-profit organization 10 0 Initial investment: 62M€ Initial staff: ~70 50 0 198 5 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 1600 1400 more than 1475 people 2006 One of the largest independent R&D 1200 551 organizations in this field, worldwide Employees 1000 2006 budget : 240M€ (includes 35 M€ grant 800 from Flanders government) 600 Staff: about 1475 400 936 Collaboration with >500 partners 200 < 15% government/state funding 0 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 FIGURE 5  IMEC: More than two decades of open innovation. PROC Figure 05 a wider range of densely integrated functionalities. We explore how to move into a multidisciplinary world where we must do many things simultaneously, leveraging on our strongly deployed research infrastructure and wide range of competencies.” Challenges for the Semiconductor Field He summarized some of the challenges for the semiconductor field in coming years. First was the cost of semiconductor R&D, which is increasing by about 12 percent annually. “This wasn’t a big deal when revenues were going up faster,” he said. “But 10 years ago revenue growth slowed, and the consensus now is for roughly 6.5 percent revenue growth. The only way to keep the R&D budget under control is by sharing costs and allowing access to external R&D.” Within the product life cycle, IMEC positions itself at the non-competitive stages, “right after university work,” where joint research is appropriate and more and more a necessity (must-have technology platforms). It acts as a “transformer” between academia and industry, providing both greater focus for universities and basic insights and solutions for industrial partners. The overall budget for IMEC was about €235 million for 2006. In 2005, the largest portion of revenues (49 percent) came from both core and non-core part- ners. Core partners included Intel, NXP, Texas Instruments, STMicroelectronics, Infineon, Micron Technologies, Samsung, Panasonic, and Taiwan Semiconductor. 

72 INNOVATIVE FLANDERS Grant of Flanders Gov’t (17.8%) Various income (2.2%) ESA (1.57%) International Industry Europactice (4.7%) (49%) EC (6.7%) Industry in Flanders (17.8%) Total revenue : 196.6 MEuro FIGURE 6  Sources of revenue in 2005. IMEC also has strategic partnerships with about06 PROC Fig a dozen equipment suppliers and many “non-core” partners. (See Figure 6.) The Flemish government and industry in Flanders each provided about €35 million, with smaller amounts from the European Commission, ESA, and others. He added that while IMEC is careful about claiming to be the biggest anything, “I think it’s fair to say that this is the world’s largest industry commitment to semiconductor research in partnership.” He showed a drawing of the IMEC campus, which holds both its original 200mm pilot line in Clean Room 1 and the new 300mm pilot line in Clean Room 2, which have total clean room space of about 8,000 square meters. About 45 tools had been installed in the 300mm room; the equipment arrived in August 2006. Toward an Interdisciplinary Future In the future, he said, IMEC would continue bringing complementary and interdisciplinary expertise under the same roof geared towards an increased speed of innovation. One example is the Neuro-electronics Convergence Laboratory. The different expertise included institutes of medicine (Leuven Faculty of Medi- cine), biology (the VIB), nano/micro electronic (IMEC), and chemistry (IMEC). All of them share facilities, space and expertise in cross-disciplinary projects at the micro- and nanoscales. Part of their philosophy was to look at many different technology options and the many trade-off aspects, because no one could tell in advance “who the winners are going to be.” He concluded by predicting that IMEC would continue to be a successful example of private-public partnership, based on opportunity seeking as well as risk taking and risk sharing. He emphasized the importance of risk taking: “If 

SESSION III: SEMICONDUCTORS 73 there is no risk,” he asked, “why would you share it?” He also expressed gratitude for the support of local government, and for the growing number of links with business partners. Discussion Dr. Spencer asked how long IMEC’s residents stay. Dr. de Proft said visits varied in length, but that he felt scientists get the most out of the experience when they stay at least 1-2 years. Dr. Wessner asked why, in view of the success of IMEC, it still received government support. The response was that IMEC needed a critical balance of fundamental and applied research, to avoid a 100 percent commercial orientation or to avoid being driven to much closer-to-the-market research—making open innovation more difficult—and that the government was the primary source of support for fundamental research. Is IMEC Subsidizing Foreign Firms? Professor Flamm asked what he termed an “impolite question” about the presence in Flanders of large multinational semiconductor companies. Was not IMEC essentially subsidizing research for these firms, none of whom had produc- tion facilities in Flanders? Dr. de Proft called the questions “astute and pertinent,” and noted that the grants were not discounts on commercial research contracts, but were meant to support fundamental research as a basis for further research programs with industry and with a view on long-term spillover effects for the region. When you dig deeper, he went on, you see payback for the region at many levels. He emphasized the presence of the residents, about 300 bright minds from around the world, spending creative years here, and building up networks. They are all people likely to move up in their organizations, where they will be in positions to make decisions about where to put their R&D centers or other activities. Furthermore, over 200 PhDs are performing their doctoral research at IMEC. IMEC also is interacting with local industry and has furthermore ­created over 25 spin-off companies, among which are some very fast growers (e.g., P ­ hotovoltech). IMEC’s activities are also generating strong economic derived impact at the region (e.g., >€42 million of subcontracting to the local industry). The overall impact, being calculated by an external expert company in 2005, is a multiple of the government funding. “Our government is smart enough to under- stand not to look for direct matches, but to promote some formative behaviors without trying to steer the economy.” 

74 INNOVATIVE FLANDERS Economic Impacts of SEMATECH on Innovation in Semiconductors Kenneth Flamm University of Texas at Austin Professor Flamm said he would discuss research that seeks to understand the past of SEMATECH, looking primarily at the 1990s, which he called “an impor- tant and dynamic period in the semiconductor world,” a time when there was increasing global dispersion of technology and production facilities. ­SEMATECH, which he called a new U.S. experiment in R&D strategy, was put in place in the late 1980s “but really drove forward in 1990s.” He said that around the mid-1990s there was a significant acceleration in semiconductor technology, when there was also a global spread of knowledge and expertise in making semiconductors. He said he would focus on microprocessors, because they were the product for which the rate of technological improvement was the fastest in the 1990s, when they were also the dominant single IT product manufactured in the United States. In 2004, almost half (46 percent) the U.S. integrated circuit (IC) shipments by value made in the United States were microprocessors, compared with 29 per- cent in 1995, and 37 percent in 2002. For DRAMs, the portion made in the United States was 14 percent in 1995 and 11 percent in 2004. Microprocessors had the highest rate of technological innovation in the 1990s, and the largest input to a PC by value. They also had a big impact on the technical improvement of com- puters and productivity in downstream IT-using industries. Finally, economists had a very rich data set on microprocessor units (MPUs) which allowed them to perform high-quality research. A Trend of Price Improvement He pointed out a notable improvement in the prices of MPUs for three periods: 1991-1995, 1995-1999, and 1991-1999. He used calculations based on common economic price index methodology and price performance improvement in different categories of semiconductor products. In each category, rates of price performance improvements (compound annual growth rates) were significantly greater for 1995-1999 than they were for 1991-1995. These data were convincing because they held not only for MPUs but were consistent “pretty much across the board. This suggests that an underlying factor was at work.” A significant part of this increased rate of decline in prices seemed to ­coincide with other events, he said. The first was a new U.S. R&D strategy, including for- mation of SEMATECH in the late 1980s. In 1992, SEMATECH sharpened its focus on manufacturing, especially to accelerate introduction of new technology nodes, using lithography as the benchmark for state of the technology. The goal was to reduce the time between new nodes from 3 years to 2, and its apparent success inspired imitation in Japan and elsewhere. 

SESSION III: SEMICONDUCTORS 75 The Role of SEMATECH An interesting feature of SEMATECH that is seldom emphasized, he said, is its coordination function. In the early 1990s, the United States developed a National Semiconductor Technology Roadmap. Begun under the aegis of the National Advisory Committee on Semiconductors at a workshop held in 1992, it evolved into a broader attempt to coordinate a complex process of technology development to a point where products would all come online when needed to advance manufacturing. The First National Technology Roadmap came out in 1994, with much of the technical leadership provided by SEMATECH. It was updated in 1997 and has been codified at 2-year intervals since. In the late 1990s, the roadmap became international and was called the International Technology Roadmap for Semiconductors (ITRS). This change recognized that semiconductor firms were now spread around the globe, and that coordination among suppliers and users had helped to accelerate innovation in the industry. The consensus is that the roadmap has helped maintain the 2-year nodes, he said, and although many people think that 2 years is not long enough to fully realize potential profits on companies’ investments, they have not been able to lengthen the cycle because of competitive pressures. This degree of R&D coordination, he suggested, was a unique structure of great interest to economists. It is the kind of activity that might invite antitrust pressure, he said, but a federal law passed in the 1980s granted limited antitrust immunity for registered consortia like SEMATECH. The international SEMATECH began in 1995 as a partnership to work on 300mm wafer technology, encouraged by the federal government. This was f ­ ollowed by the recovery and stabilization of the U.S. semiconductor industry. The U.S. government subsidy ended in 1997, and today the share of world semi- conductor output accounted for by SEMATECH members exceeds the share when it was formed in late 1980s. In September 2004, the “international” designation, too, was dropped, though the organization still has many full international members; the most recent to join is Samsung. It has spun off a subset of R&D activities into the International Semiconductor Manufacturing Initiative (ISMI), which walled off access to the “highest tech” activities (e.g., lithography). The main SEMATECH organization has nine “full” members (AMD, Freescale, Hewlett-Packard, IBM, Infineon, Intel, Phillips, Samsung, and Texas Instruments), who also have membership in ISMI. It also has three ISMI-only members who do not get access to full S ­ EMATECH information: TSMC, Panasonic/Matsushita Electric, and Spansion. The first ­Japanese member was Renesas, followed by NEC in 2006. But even as SEMATECH went international, the U.S. semiconductor indus- try’s global share of R&D declined as U.S. firms moved more functions offshore. In the 1990s, there was resurgence of semiconductor leadership in U.S. after some years of decline, and U.S. semiconductor firms again moved to the top. Then R&D coordination through the roadmap in the 1990s brought coordination with 

76 INNOVATIVE FLANDERS suppliers in areas where the “best of breed,” he said, were no longer located in the United States. After the millennium, increasing offshore competence again led to some increase in offshoring in R&D by U.S. firms. New Models of R&D Coordination Since then, he said, interesting new models of R&D coordination have emerged—largely because semiconductors now require very large investments in R&D. One model, subsidized by and located in New York State, is a hub- and-spoke system. IBM, the hub, works with three core partners in developing manufacturing process technology: Samsung, Infineon, and Chartered. Toshiba and Sony are also involved, as is AMD. The model is probably somewhat less open than IMEC or SEMATECH because the partnership is negotiated one on one with other core members. Another partnership is Crolles II, formed by Phillips, STM, and Freescale. This group also has government support and international composition. What are the benefits of such models? he asked. An obvious one is some acceleration in the rate of manufacturing innovation, such as the new 2-year technology nodes begun in the 1990s. Benefits of Shorter Times Between Nodes Another benefit, related to the use of roadmaps, is improvements in price performance, which may be viewed in two ways. One is by engineering efficien- cies in products already made—lower price for a given quality or functionality. A second is new capabilities that become possible because of pooled technology. These benefits, he said, are not independent. By shrinking the features of a chip, a company can not only produce the same chip in a smaller area, thereby saving cost, but also the chip can be faster. So while shrinking the technology nodes from 3 years to 2 years gained about 50 percent in price performance, this gain had two components. Roughly half the decline was due to improvement in processor quality, but acceleration in technology nodes also led to acceleration in processor speed. This is because a byproduct of smaller feature sizes is shorter distances between features, which allows for faster chips. Design innovation is needed to make use of greater switching speeds, which is a big factor in user evaluation of processor quality. So this gain in speed, he concluded, was another benefit of the acceleration in nodes, beyond merely reducing manufacturing cost. The Importance of Manufacturing Gains Price performance improvement, however, had slowed in the last year and a half as MPUs hit a “brick wall” related to power and heat dissipation in 2004- 2005, and this decline coincided with a slowdown in the rate of processor speed 

SESSION III: SEMICONDUCTORS 77 increases. A much bigger share of price performance was now due to improve- ments in manufacturing costs. He concluded by saying that the R&D coordination that began with S ­ EMATECH and continued through the years of the international roadmap appeared to have created significant economic benefits over the last decade. More recently, the gains made in the manufacturing process have become more impor- tant because the rate of improvement in other key components has slowed. IMEC and SEMATECH: An Industrial Partner Perspective Allen Bowling Texas Instruments Dr. Bowling, who manages Texas Instruments’ external research activities in silicon technology development, said that in the 2000s his company had “moved from an era of microelectronics to nanoelectronics,” routinely producing gates as small as 40 to 50nm. He said that he would talk about where this would lead in the future as the trend toward nanoelectronics continued. More Dependence on Consortia and Outside Knowledge At the fabrication facilities of Texas Instruments’ Dallas headquarters, the company had adopted the roadmap with 2 years between nodes described by Professor Flamm. Because it takes at least 4 years to fully develop each node, said Dr. Bowling, the company has two to three of them in co-development at any time. In-house technology development programs start about 3 years before manufacturing, and the development group is involved until about a year after manufacturing begins, which amounts to a 4-year period. Altogether it takes about 7 to 12 years to move a new material or device into production. This means that its engineers are more dependent on the long-range knowledge resources of SEMATECH, IMEC, and universities to keep up with current fundamental and applied research and work at the current accelerated pace. They also depend on close collaboration with equipment suppliers. Texas Instruments has other ways of staying current with technology devel- opments. Beginning about 5-20 years before product qualification, Texas Instru- ments collaborates more than it ever has with universities, especially through the SRC consortium. Texas Instruments provides direct funding for some short- term SRC needs and is a member of the three SRC consortia. Altogether, these consortia are supported by about $55 million per year from industry and about $20 million per year from the federal government, and other public funding. The program provides funding for about 1,000 graduate students; Texas Instruments supports student preparation generally, on the premise that students represent the future employees for the industry. 

78 INNOVATIVE FLANDERS The company also promotes partnerships with states to support research infrastructure. He stressed the importance of local government involvement in supporting facilities like IMEC, including a location and a fab structure. State support allows the companies in consortia to focus on running programs, which is the strength of industry. The reason the IBM-centered consortium is viable in New York State is that the state is funding the infrastructure, as does Flanders for the IMEC facility. Why Texas Instruments Belongs to Both IMEC and SEMATECH Texas Instruments has been a charter core member of SEMATECH since it began in 1987; there are currently eight core members. It has also been a core member of IMEC since 2004, after following activities in selected programs since 1993. Texas Instruments is willing to pay its membership dues in both SEMATECH and IMEC because, said Dr. Bowling, “it earns a high return on its investment.” Both programs spend over $100 million per year on the pre- c ­ ompetitive research needs of the next one to two nodes, and Texas Instruments is able to leverage the knowledge it gains from association with the eight or nine other members. They belong to both consortia because each has unique capabilities. S ­ EMATECH is best at driving the international roadmap. Members gather to discuss the issues “so the right attention goes to the key gaps for the future.” It is also a true industry consortium run by members who each have one equal vote to determine exactly what kind of instrument to work on. The voting can be a rallying point when there is agreement, or a problem when member companies do not agree. Texas Instruments also belongs to IMEC, where it finds several advantages: • Advanced equipment. IMEC has advanced immersion 193nm and EUV lithography, through close alliance with ASML, which is based in the ­Netherlands. Their equipment is “above that of any other consortium in the world.” • Development collaboration. IMEC has provided much more develop- ment collaboration for equipment suppliers. They can tailor individual relation- ships with suppliers, agreeing to work on co-development and keeping the results reasonably confidential until the capability is proven. After that, they begin shar- ing it with the other member companies. • Focus on fundamental science. IMEC focuses more on the fundamentals of why things happen. Many research staff are also university faculty, and there are many grad students working at IMEC with strength in advanced device con- cepts demonstration and testing. • Public support. Public support from Flanders, and smaller amounts from the EU, allow IMEC to keep its research infrastructure at the state of the art, which is critical for the competitive position of global companies. 

SESSION III: SEMICONDUCTORS 79 He also listed several disadvantages of the IMEC system: • Less focus on manufacturability issues. While IMEC’s strength lies in advanced device concepts demonstration, it focuses less on manufacturing. • Less management control. Even though it has core members, manage- ment reserves right to determine key areas to work on. This means that not all of a member’s key issues will be addressed. But IMEC is good at focusing on key issues when diverse opinions exist. He summarized the advantages of consortia with a sobering statistic. A typical integrated circuit manufacturing flow involves more than 1,000 process steps. The complexity of this flow provides numerous opportunities for col- laboration on many challenges, and provides strong rationales for the roles of both ­SEMATECH and IMEC. Texas Instruments receives unique value from its memberships in both IMEC and SMT, he said, because they each focus on areas of particular strength.