Fundamental physical limitations in semiconductor scaling have slowed future expectations of improved, single-processor computing performance on which all sectors of society in the United States—and around the world—have relied. Many of these technological challenges at the frontiers of device design, computer architecture, and parallel programming methodologies were described in Chapter 1. While many short-term technological fixes have led to recent computing performance improvements, no silver bullets have emerged to reclaim the steady exponential computing performance gains once achieved by successive generations of single microprocessor computer systems.
How, then, will the next generation of semiconductor, computer architecture and programming breakthroughs come about? What types of policies and institutions, whether public, private or partnerships, will be required to bring about the technological innovation necessary for next-generation hardware devices, system architectures, and programming systems? To address these questions, it is important to understand the role innovation policies have played in supporting U.S. innovation and, in the context of this report, the policies that strengthen, sustain, and/or erode the innovative capabilities of the computer and semiconductor industries.
Innovation policies differ across countries, industries, and technologies. Countries differ in their levels of development and in their economic institutions, and hence pursue quite different approaches to innovation policy. The United States model, for instance, has largely been based on a belief that market forces (which include government and defense as consumers that demand leading-edge technology) and the private sector should play a primary role in innovation, backed by government investment in basic academic research. In contrast, the European Union and emerging economies such as China, Korea, and Taiwan rely much more on the government to articulate strategic objectives and key parameters.
In the United States, there is a widespread expectation that government-centered innovation systems will “naturally” converge to a U.S.-style market-led system. However, comparative research on national innovation policies suggests that this convergence is limited.1 In addition, innovation policies change over time, even within the same country. As Charles Vest emphasizes, the American innovation system has a long tradition of highly decentralized, market-driven innovation networks, where the government historically played a role primarily at the local level.2 However, as ubiquitous globalization disaggregates manufacturing, product development, and research, it is not yet clear which policies will best support future innovation in the United States.3
To understand how requirements for innovation policy differ across industries and technologies, it is
1See R. R. Nelson, 1993, ed., National Innovation Systems. A Comparative Analysis, Oxford University Press, New York. For an analysis of the persistent diversity of China’s and America’s innovation and standards policies, see D. Ernst, 2011, Indigenous Innovation and Globalization: The Challenge for China’s Standardization Strategy, University of California Institute on Global Conflict and Cooperation, La Jolla, CA, and East-West Center, Honolulu, HI, 123 pp.
2C. Vest, 2011, “Universities and the U.S. Innovation System,” in C. W. Wessner, ed., Building the 21st Century. U.S.-China Cooperation on Science, Technology, and Innovation, Washington, D.C.: The National Academies Press, pp. 70-73.
3D. C. Mowery, 2009, Plus ca change: Industrial R&D in the “third industrial revolution,” Industrial and Corporate Change, (18):1, pp. 1-50.
useful to consider how innovations differ—in the complexity of the infrastructure and capabilities required to foster and implement them. Furthermore, the demands of innovation policy differ, depending on the nature and the intensity of innovation barriers that constrain the deployment of new ideas, inventions, and discoveries into commercially successful products, services, and business models.
Today, the effects of globalization extend across all stages of the value chain, including engineering, product development, and applied and basic research. This has resulted in an increase in the organizational and geographic mobility of knowledge.4 However, the new geography of knowledge is not a flatter world where technical change and liberalization spread the benefits of globalization rapidly and equally. Instead, even mature and established technology and manufacturing leaders now face competition from a handful of new—yet very diverse and intensely competitive—manufacturing and research and development hubs around the world.5 Therefore, the United States can learn a great deal by looking at the strengths and weaknesses of alternative information technology (IT) innovation policies in other nations.6 An analysis of these diverse approaches to innovation policy is shaped by issues such as: the range of policy options that have been pursued, how policy approaches differ, how these differences affect innovation capacities, and how innovation policies pursued elsewhere affect the global supply chain.
This chapter examines the strengths and weaknesses of different innovation strategies, policy tools, and institutional arrangements implemented in countries that are potentially important players in the development of computing devices, technologies, and products. While U.S. innovation strategies have primarily relied on market forces and the private sector, it is important to understand the varied and complex factors that drive the evolution of different national innovation ecosystems. For example, countries such as China and Taiwan have relied on top-down government leadership to define strategic objectives and key parameters of innovation programs. Another variant of innovation policy can be found in the European Union’s recent push toward new forms of cross-border coordination of innovation markets and infrastructures.
Section 3.1 provides a history of the U.S. semiconductor industry and examines how America’s decentralized market-driven innovation system has led to where the United States is today. Section 3.2 looks at China’s indigenous innovation policy, especially its recent Strategic Emerging Industries (SEI) Program. Section 3.3 examines the evolving role of Taiwanese innovation policies to support low-cost and fast innovation through domestic and global innovation networks. Section 3.4 looks at Korea’s coevolution of international and domestic knowledge linkages. Section 3.5 examines the European Union’s recent efforts to develop an integrated innovation strategy and its recent Key Enabling Technologies (KETs) Program. Section 3.6 provides concluding remarks and policy implications.
3.1.1 Historical Context
Several factors influence the range and type of policy options available to nations to promote and manage development and competitiveness in the semiconductor, computer architecture, and software programming arenas. Among those factors historically dominating U.S. policy considerations are
- The economic importance of semiconductors and computing in the U.S. national economy;
- The economic importance of closely related U.S. industries (e.g., telecommunications, consumer electronics, military and aerospace);
- The outlook on the U.S. federal budget, the climate for public and private investment, the employment picture, and predictions on economic growth;
- Political perceptions about the health of these industries relative to others;
- Public perceptions about the United State’s competitive commercial position, as well as leadership of the United States vs. other nations, in these industries;
- Both real and perceived dependence of U.S. intelligence and national security on leadership in these industries, and U.S. reliance on foreign technologies and assistance in areas related to intelligence and national security; and
- Prevailing political philosophies regarding industrial policy.
4D. Ernst, 2005, “The New Mobility of Knowledge: Digital Information Systems and Global Flagship Networks,” in R. Latham and S. Sassen (eds.), Digital Formations: IT and New Architectures in the Global Realm, Princeton University Press, Princeton and Oxford.
5D. Ernst, A New Geography of Knowledge in the Electronics Industry? Asia’s Role in Global Innovation Networks, Policy Studies, No. 54, August 2009, East-West Center, Honolulu, HI, 65
6This is in line with Jacques Gansler’s argument for a “global strategy” made for the U.S. defense industry (J. Gansler, 2011, Democracy’s Arsenal: Creating a Twenty-First-Century Defense Industry, The MIT Press, Cambridge, MA).
The last two decades have exposed significant conflicts among these traditional influences on policy, largely brought about by three important changes: (1) the end of the cold war, (2) the general stagnation of the Japanese economy, and (3) the globalization of the computer and semiconductor industries into well-established7 and mutually dependent supply chains and markets.
Federal funding of electronic development, from the launch of Sputnik in 1957 almost until the fall of the Berlin Wall, was driven by perceived military requirements, which had significant noneconomic motivation. During this period, significant federal R&D investment was made in innovative semiconductor technology for military application. After some cost reduction and normal technology adoption delay, the same technology and technology roadmap steadily appeared in the commercial market, including advanced compound semiconductors and dramatically new manufacturing equipment, which also found strong commercial adoption, for example, in lithography.
By the mid-1990s this pattern had reversed; that is, the incredible acceleration of the personal computer (PC) and server industries meant that commercial technology was leading rather than lagging behind military technology. This shift led to an increasing focus on the use and adaptation of commercial off-the-shelf technology in federal procurement and contributed to the steady decline in federal funding for R&D,8 given the U.S. preeminence in the area and the already high levels of research investments by the U.S. computer and semiconductor industry.
Today, cutting-edge R&D in semiconductors, the historical engine of computer performance growth, has become unmanageably expensive for the usual U.S. federal agencies. At the same time, it is extremely difficult for industry to invest in long-term R&D, given the near-term expectations of the financial markets. One consequence has been limited commercial R&D investment in hardware and software technologies whose economic return is not realized rather quickly.
In the United States, industrial policy has typically not been viewed as an offensive tool for economic competition or a means to create new industries or accelerate successful ones, but rather as a defensive tool to protect or restore existing industries under competitive economic pressure.9 The perception of favoring certain industries, “picking winners,” by government pressures and incentives, rather than allowing for natural market forces and laissez-faire investment, has been politically toxic. On the other hand, rescuing at least some foundering industries, or attempting to regain lost ground in critical ones, has been generally politically rewarding. The Asian competitor nations, for example, China and Japan, traditionally have both subsidized and protected (by legal and covert subsidies and tariffs) those industries that they choose to target.
In contrast, it is important to recognize that U.S. industries, and information technology in particular, do not tend to receive attention or assistance from federal sources simply because they are slowing down in growth or maturing; there typically must be a specific adversary. For example, once U.S. superiority in electronics and computation (e.g., for guidance systems) over the Soviet Union became assured, the focus of government policy switched to the rising Japanese dominance in electronics, especially including memories.
While countries such as Japan began forming R&D consortia as early as 1956, the practice was illegal in the United States until Congress passed the National Cooperative Research Act in 1984.10 Two years later, concerns of a U.S. decline in semiconductor market share prompted a call by the Semiconductor Research Corporation (SRC) and Semiconductor Industry Association (SIA)11 for increased cooperation to provide the U.S. semiconductor industry with the capability of regaining world leadership in semiconductor manufacturing. As a result of this effort, SEMATECH (Semiconductor Manufacturing Technology) was created in 1987 as a partnership of 14 U.S. semiconductor companies with the Defense Advanced Research Projects Agency (DARPA), which contributed U.S. $500
7While well established and interdependent, these value chains can be highly vulnerable to sudden disruptions from natural disasters, geopolitical conflicts, and so on. Some of these are discussed in greater detail in Chapter 4.
8J. Gansler, 2011, Democracy’s Arsenal: Creating a Twenty-First-Century Defense Industry, The MIT Press, Cambridge, MA.
9U.S. innovation policy can be thought of as “market conforming” in its intent to address problems that economists have deemed weaknesses for technological advancements. In particular, these were externality problems that required collective R&D funding and that funding took specific paths because of appropriation processes in Congress.
10D. V. Gibson, and E. M. Rogers, 1994, R&D Collaborations on Trial, Harvard Business School Press, Cambridge, MA.
11“Founded in 1977 by five microelectronics pioneers, SIA unites over 60 companies that account for 80 percent of the semiconductor production of this country.” (see www.sia-online.org) SIA, along with the European Semiconductor Industry Association (ESIA), the Japan Electronics and Information Technology Industries Association (JEITA), the Korea Semiconductor Industry Association (KSIA) and the Taiwan Semiconductor Industry Association (TSIA), sponsors the International Technology Roadmap for Semiconductors, a15-year assessment of the semiconductor industry’s future technology requirements (see www.public.itrs.net). Last accessed on June 30, 2012.
million over 5 years, to solve common manufacturing problems and to regain U.S. competitiveness in the semiconductor industry that had been lost to Japanese industry in the mid-1980s12. In the committee’s view, SEMATECH played a strong role in early efforts to reclaim U.S. semiconductor manufacturing leadership and has been a successful example of a U.S. consortium demonstrating the value of federal funds and federal participation. This position is reiterated by a 2002 National Research Council report, Government-Industry Partnerships for the Development of New Technologies, which found that the SEMATECH partnership directly contributed to the global competitiveness of U.S. industry, specifically the resurgence of the U.S. semiconductor industry.13
Today, the SRC also continues to play a significant role in advancing the semiconductor industry though synergetic industry and university research programs and support initiatives around the world, such as the Global Research Collaboration Program, Nanoelectronics Research Initiative, Focus Center Research Program, and Semiconductor Research Corporation Education Alliance. The National Nanofabrication Infrastructure Network (NNIN) also provides a successful example of U.S. government (National Science Foundation) support of university semiconductor research. By paying for some expensive semiconductor research equipment at universities, the NNIN enables leading-edge research, which indirectly supports the U.S. semiconductor industry with research results and science and engineering graduates.
In contrast, two other industry-only consortia, started near the same time and for similar reasons, both failed. The Microelectronics and Computer Technology Consortium (MCC) was formed in the early 1980s as a response to Japan’s Fifth Generation Computer Systems (FGCS) project.14 Entirely funded by corporate partners, MCC worked on a wide range of technology and software projects, with early sponsorship particularly from mainframe computer companies. By 2000 the Board of Directors had decided to dissolve the organization. Another industry-only consortium, U.S. Memories was organized in 1989 to manufacture memories based on technology from IBM, to avoid dependence on Japanese vendors. However, by early 1990 the consortium members had proven unwilling to make the necessary investments, and major memory users like Apple, HP, and Sun did not participate, so the project was canceled. Thus far, consortia that include IT competitors but that do not have government leadership have fared poorly, due to a combination of mutual suspicion, lack of focus, and no real sense of urgency.
In summary, U.S. federal support and investment has historically relied upon a perception of military threat, economic decline, industry crisis, and/or loss of competitive position; and in the United States, electronic and computer consortia without both federal R&D support and federal direction have not generally succeeded. Thus, centralist technology policies that may work in nations and cultures that accept such direction readily are a poor match to the U.S. free-market model. Further, innovation policy has to reflect each country’s unique economic institutions, industry structure, and growth model.
3.1.2 Global Semiconductor Competition
While it could be proposed that some U.S. computer vendors “failed to innovate,” or “gave up the fight” to foreign competition, it is important to recognize the paired advantages and shortcomings of a free-market industrial economy, and the capacity it provides for innovation, not only in technology, but in the creation (and destruction) of whole economic sectors. U.S. capital market investors are often quick to spot and to capitalize on transformative shifts in a business paradigm, and, consequently, to move their investments in a way that often accelerates the change. Capital markets tend to value short-term quarterly profits and tend to reward or punish a company accordingly, which manifests in changes in its stock price. This has advantages and disadvantages. On the one hand, it discourages waste and encourages competitiveness. On the other hand, a short-term focus often discourages long-term thinking and R&D investment, particularly during difficult economic times. Federal R&D programs and public-private consortia play a crucial role in coping with this tension.
In the late 1980s, when it appeared that focused government programs in Japan, as well as unfair or unreasonable trade practices, might overtake U.S. competitiveness, DARPA investments, especially SEMATECH, drove the necessary R&D efforts in process and equipment to sustain Moore’s Law and to maintain the confidence of capital markets. Concurrently, IBM began to accelerate its investment in very high-performance semiconductor technology and to form joint innovation partnerships with numerous (non-
12See www.sematech.org/corporate history; www.sematech.org/corporate/timeline; NRC, 2003, Securing the Future: Support to the Semiconductor Industry, Washington, D.C.: The National Academies Press (available online at http://www.nap.edu/catalog.php?record_id=10677#toc).
13NRC, 2002, Government-Industry Partnerships for the Development of New Technologies, Washington, D.C.: The National Academies Press (available online at http://www.nap.edu/catalog.php?record_id=10584).
14Kazuhiro Fuchi, 1984, Revisiting Original Philosophy of Fifth Generation Computer Systems Project, FGCS, pp. 1-2.
Intel) semiconductor fabrication companies, creating a business counterpoint to Intel. More recently, the rise of mobile computing devices has created new competition, both among existing companies and new ones formed in response to emerging market economies.
3.1.3 Creation of the U.S. (and Global) Software Industry15
Continuous technical innovation that sustained Moore’s Law (and exponential growth in computing performance) led also to the creation of the commercial software industry as a meaningful force in the U.S. economy. This took place in two ways. First, the falling cost and wide availability of powerful microprocessors greatly increased the number of computers in use, and successful software products could be sold in enormous numbers at modest prices.
Second, the fact that a small number of instruction set architectures (ISAs) dominated the PC and server marketplace16 meant there was a larger and consolidated software market that would benefit from steady improvements in cost and performance, while seldom requiring any significant changes to the programs. Vendors rarely prefer to use new instructions sets until they have been in the market for many years and are available on a significant fraction of deployed machines. This allowed larger software investments to be made, in products that would surely perform better over time, courtesy of Moore’s Law.
U.S. firms dominate this 30-year-old PC and server industry, although a few European firms (e.g., SAP) are of significant size and share of the market. There is early evidence that this market dominance may extend to the new world of smartphones and tablets, as well as cloud services, though global competition in this space is new and intense.
3.1.4 Consequences of the U.S. Free-Market Approach
These three phenomena—the enormous growth of the semiconductor industry, the commoditization of the computer industry, and the emergence of a huge software industry—are mutually dependent, and have created a virtuous economic framework. They also afforded the United States the opportunity to achieve and maintain its leadership in information technology generally. Although federal support to long-term R&D has been indispensible, particularly in bad economic environments, such achievements would almost certainly not have been possible under a centrally managed policy regime. For instance, in a more managed environment, the policy impulse might have been to save legacy computer companies; instead, market forces coupled with a noninterventionist approach and (to a lesser extent) government antitrust efforts helped ensure the “creative destruction” that has transformed computing and the role of the United States in it. The United States has been rewarded by the emergence of very strong semiconductor design and software industry leadership—in exchange for the loss of some semiconductor fabrication and the assembly and testing of commodity products to foreign vendors, for example, the off-shore assembly by contract manufacturers of even the strongest U.S. computer brands, based upon cost.
However, staying strictly on any technical path involves bypassing others, and sacrificing progress in some areas to sustain others. It is certainly worth examining some of the approaches delayed or abandoned by the course taken by the IT industry.
In the 1960s and 1970s, computers were expensive, resources were limited, and programmers were scarce. There was great emphasis on creating clever algorithms that required the least number of instructions or smallest amount of memory, or both. Elegant, parsimonious program design was celebrated, and improvements to compilers for denser code and new languages for programmer productivity were high priorities in academia and industry alike. High-productivity programming languages help programmers produce working programmers faster, as compared with high-performance programming languages, which help programmers extract as much performance as possible by exposing machine details to them. From high-productivity languages and the relentless hardware performance improvements enabled by Moore’s Law, a new and much larger pool of programmers emerged. These programmers applied application-specific knowledge, for example, machine learning, graphics, animation, accounting, government functions, and so on, driving an explosion in software capabilities in the era of ever-faster and cheaper central processing units (CPUs) and memory. In addition, programming emphasis moved from performance productivity to getting new products out faster.
Another area of technology research and innovation affected by Moore’s Law was parallel computing. In the 1970s, very large scientific computers with parallelism among several arithmetic units were just beginning to
15Section 3.1.3 and part of Section 3.1.4 rely heavily on David Liddle, “The Wider Impact of Moore’s Law,” IEEE SSCS Newsletter, September 2006. Available at http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=4785858. Last accessed August 21, 2012.
16The history of the mobile and embedded computing space is much more varied, with a diversity of ISAs and vendors.
work well. Equally important, software researchers were beginning to make real progress on the problem of programming systems for parallel machines. Moore’s Law advances relegated most work in parallel computing to business servers and scientific and technical computing.17 A revival of parallel computing research and development in the 1990s yielded several new approaches and companies, but the early promise was not realized, for many reasons. Because the size of the technical computing market was small relative to the PC market, the research and product development took a different path, focusing on performance maximization at reduced costs rather than ease of use and programmer productivity.
Only now that the limits of growth in CPU clock frequency are in sight for consumer devices has serious focus on parallel processing reentered the mainstream. Earlier research from the 1980s and 1990s showed the difficulty of developing tools that can easily convert legacy sequential codes into scalable parallel code that run well on current generation or next-generation machines. This research experience suggests that future work should emphasize simplicity and programmability of heterogeneous multicore devices to address mainstream product needs. In addition, such research should be driven by an awareness of the demonstrated limitations of automatic parallelization and recognition that the intrinsic parallelism in application problems differs markedly.
Another effect of Moore’s Law over the years has been that the capital markets, given the visible, vast investment in scalable complementary-symmetry metal-oxide-semiconductors (CMOS) (except for extremely low-volume exotic noncommercial uses) have not encouraged R&D on new post-silicon materials. Even small deviations from the CMOS path, like silicon-germanium or silicon carbide have aroused skepticism, let alone compound III–V semiconductors.
Typically, a free-market approach will continue on a profitable path until it begins to reach diminishing returns; unfortunately, that point is sometimes recognized too late. The uniprocessor CMOS clock-frequency race has already ended, leaving the United States ill prepared with either semiconductor or software succession plans. This type of situation has traditionally been one in which U.S. government participation with academia and industry has been effective. Neither the U.S. federal government nor the U.S. computer industry has come to consensus on a strategy that ensures U.S. leadership in the next generation of computing technologies. Thus, those developing future U.S. policy in these areas should carefully consider the opportunities and consequences of alternative innovation strategies, some of which have been tried elsewhere.18
Over the last several decades, China has made significant efforts to align its science, technology, and innovation policies to support indigenous innovation. These trends towards technonationalism19 were prompted by political concerns within China that it both lacks indigenous technology and depends on foreign technology, as well as from several lessons learned over the past several decades. For example, after its relationship with the former Soviet Union ruptured in the 1950s, China shed its reliance on Soviet technology and developed a national strategic weapons program, developing its own nuclear weapons, missiles, and satellites.
Then, in the late 1970s, China embraced globalization by opening its huge market to multinational corporations for the exchange of advanced technology. In turn, the rapid growth of China’s semiconductor consumption primarily reflects its emergence as the dominant global factory for IT equipment. Between 2004 and 2009, alone, China’s share of global electronic equipment production increased from 17 percent to 31
17It is worth noting that parallel computing work continued in the high-performance computing (HPC) and server segments.
18When examining innovation strategies elsewhere, it is important to recognize that most are being applied to those whose attempts to enter the advanced computing market were late relative to the United States. Latecomers have disadvantages and advantages that need to be assessed and taken into account when considering policy options and what lessons might be learned. For instance, latecomers to advanced computing need to overcome very substantial entry barriers (disadvantages), as well as to exploit new opportunities that result from beginning with less-complicated technology and having fewer legacy constraints on technology development, strategy, and organization (advantages). Economies of scale may be a barrier to market entry requiring nonprice means of market penetration, that is, through product differentiation and the creation of new markets and distribution channels. On the other hand, latecomers who become fast followers of established technology roadmaps are able to set clear targets for product development and related research, as well as to compare and learn from the experience and failures of incumbent leaders. Latecomers are also not locked into supporting and maintaining legacy technologies or infrastructures.
19Policy orientation towards autonomy and independence from other states (see B. Naughton and A. Segal, 2001, “Technology Development in the New Millenium [sic]: China in Search of a Workable Model,” MIT Japan Program, Working Paper Series 01.03., May 28).
percent.20 This marked increase suggests a reshaping of the IT equipment manufacturing landscape.
However, as the global financial and economic crisis continues, exports from China have slowed, placing pressure on China’s export-oriented economic development model. In pursuit of new growth engines for its economy, several government policies and initiatives have facilitated the strengthening Chinese indigenous innovation.
3.2.1 Government Policies and Initiatives to Strengthen Indigenous Innovation
Medium and Long-Term Plan (2006–2020)
In early 2006, China released its Medium- and Long-Term Plan for the Development of Science and Technology (MLP) (2006–2020). This plan set the tone for strengthening China’s indigenous innovation capability by addressing four problems in China’s scientific and technological development: (1) lack of innovation in commercial technologies and dependence on foreign technology; (2) increasingly unfriendly international environment for acquisition of foreign technologies; (3) technological failure to meet critical energy, water and resource utilization generally, and environmental protection and public health needs; and (4) mounting technological challenges for meeting national defense needs.21 While not specifying what indigenous innovation means, the MLP highlights three channels through which indigenous innovation capabilities may be strengthened: (1) genuine “original innovation,” (2) “integrated innovation” (fusing together existing technologies in new ways), and (3) “reinnovation” (assimilation and improvement of imported technologies).
The MLP singles out 16 engineering mega-programs, as well as identifies 11 key areas, 8 frontier technologies, and 4 science mega-programs, to support in the next 15 years (see Appendix I). Many of these programs and focus areas are directly relevant to advanced computing, including the IT industry and modern services (key areas); information technology and new and advanced materials (frontier technologies); and core electronic components, high-end generic chips, basic software, extra-large-scale integrated circuit (IC) manufacturing and techniques, and new-generation broadband wireless mobile telecommunications (megaengineering programs).22
Strategic Emerging Industries (SEI) Program
Strategic emerging industries (SEIs) refer to industries23 associated with the development of technologies (e.g., information, biotechnology, medical, new energy, environment, marine, and space) that have strategic importance to China; many are similar to the frontier technologies prioritized in the MLP. These SEIs have been said to represent the future direction of industrial development in China and will play a critical role in its continuous and sustainable economic growth, particularly in national economic and social development and optimization and upgrading of industrial structure.
Launched in October 2010,24 the SEI Program was highlighted as an important component of the 12th Five-Year Plan for National Economic and Social Development (2011–2015). As selected SEIs are science and technology based, the SEI Program is expected to decrease China’s dependency upon external technology and boost indigenous innovation capabilities, ultimately spurring economic growth and the formation of a new industrial cycle. It is expected that the government will also work out financial and taxation policies to support, guide, and encourage capital investment, and establish special funds for the development of SEIs.
Foreign-invested design subsidiaries of leading foreign semiconductor companies and global original equipment manufacturers (OEMs) play an important role in China’s chip design industry. Of the 472 design enterprises reported in China at the end of 2009, approximately 100 were the design units or activities of foreign-invested or subsidiary multinational
20PwC, 2011, Continued growth: China’s impact on the semiconductor industry – 2011 update, p 13. Available at www.pwc.com/gx/en/technology/assets/china-semiconductor-report-2011. pdf. Last accessed January 27, 2012.
21C. Cao, R. P. Suttmeier, D. F. Simon, 2009, “Success in State Directed Innovation? Perspectives on China’s Plan for the Development of Science and Technology,” in G. Parayil and A. P. D’Costa (eds.), The New Asian Innovation Dynamics: China and India in Perspective, Palgrave Macmillan, London, pp. 247–264.
23SEIs include the following: Energy-saving and environmental protection, new generation of IT, biotechnology, and high-end equipment manufacturing industries; new energy, new materials, and new energy automobile industries. State Council of China, 2010, Decisions of State Council on Accelerating the Cultivation and Development of Emerging Strategic Industries, G.F. No.32, October 29 [USITO Draft Translation].
24On October 18, 2010, the State Council issued a “Decision on the Acceleration of Nurturing and Developing Strategic Emerging Industries,” formally launching the SEIs. The 12th Five-Year Plan for National Economic and Social Development (2011–2015), released in 2011, includes SEIs as one of its important components.
companies.25 These foreign-invested design subsidiaries engage in a variety of activities that range from the simple to the complex, including adapting parent company product standards for the China market, providing lower-cost capacity for standardized back-end design functions that are integrated in the parent company’s design flow, but they also include integrated design projects for system-on-a-chip (SoC) designs.
In addition to multinational partnerships, long-term investments in the IT and advanced computing industries are now beginning to yield competitive, indigenous microprocessors, designed wholly in China. Over the last decade, the Chinese Academy of Sciences (CAS) has been developing and prototyping their line of Loongson and Godson processors.26,27 During this time, each generation of Loongson or Godson processor has become more capable and is now rivaling leading-edge microprocessors in performance. Similarly, the ShenWei series of microprocessors,28 developed by the Jinan Institute of Computing Technology (affiliated with the People’s Liberation Army) since 2006, has found its way into the first home-grown supercomputer in China, called BlueLight. As of November 2011, BlueLight includes 8,704 ShenWei chips and achieves Linpack performance of 795 TFlops,29 while being one of the most energy-efficient computers in the world.
While neither of these programs has shown any commercial success either inside or outside of China, the investment period is long and ongoing. One would expect that such programs aspire to serve the Chinese domestic market, at least for high-end technical computing. As these, and other, systems diverge from the x86 and ARM ecosystems, they will have different sets of innovation and performance optimization challenges. It is likely that national innovation policies, both in China and in the United States, will have to be iteratively redefined to meet these challenges. This is also the case should Chinese-designed computing systems begin to have any commercial success inside or outside of China.
It is worth noting that China does not have any representation among the largest semiconductor companies, nor does it have any mass commercial processors or architectures. China does, however, have a seat at the six-seat World Semiconductor Council. This seat may reflect both China’s growing potential as a semiconductor market and its increasing capabilities in semiconductor design. China’s increasing share of worldwide patents focused on semiconductor technology, from 13 percent to 22 percent in 2009, also suggests improvements in China’s semiconductor innovative capacity. More significantly, China’s share of semiconductor patents that are first issued in China has grown from 0 percent in 2005 to ~24 percent in 2009.
3.2.2 Impacts of Government Policy Efforts
It is unclear to what degree the MLP will enhance China’s indigenous innovation capability. However, international technical and business communities have already expressed concern over Chinese efforts to support the indigenous innovation efforts. For example, government policies, which gave preference for procuring domestic technologies and products in the name of supporting indigenous innovation, were abolished, at least temporarily, due to extreme pressure from foreign governments and companies. In addition, some studies30 have characterized China’s innovation policies as a threat to global intellectual property rights; a recent report by the U.S. Chamber of Commerce has even claimed that Chinese innovation policy is “a blueprint for technology theft on a scale the world has never seen before.”31
Lastly, the committee believes, growth in China’s homegrown industrial capacity, plus China’s massive urbanization, has nurtured an increasingly large domestic market in different manufacturing—and increasingly R&D—sectors. It is not clear to what degree other key
25PwC, 2011, Continued growth: China’s impact on the semiconductor industry – 2011 update. Available at www.pwc.com/gx/en/technology/assets/china-semiconductor-report-2011.pdf. Last accessed January 27, 2012.
26Godson 3B is an 8-core processor with vector extensions implemented in a 65 nm technology. It employs a MIPS instruction set (originally developed in the United States), runs at 1 GHz, and has a peak performance of 128 GF.
27The Loongson and Godson processors were developed under the leadership of U.S.-trained computer scientist Li Guojie at the Chinese Academy of Sciences’ Institute of Computing Technology, CAS.
28The ShenWei 3 chip contains 16 cores, is implemented in 65nm and runs at 1.1GHz.
29There are two other Chinese HPC installations above BlueLight (no. 14) in the November 2011 TOP500 List. However, both of these are built from Intel CPUs and NVIDIA graphics-processing units, delivering 2.566 PFlops (no. 2) and 1.271 PFlops (no. 4). The BlueLight cluster is ranked no. 39 on the Green500 List of the most efficient supercomputer clusters.
30U.S. International Trade Commission, 2010, China: Intellectual Property Infringement, Indigenous Innovation Policies, and Frameworks for Measuring the Effects on the U.S. Economy, November.
31James McGregor, 2010, China’s Drive for “Indigenous Innovation”: A Web of Industrial Policies, Global Intellectual Property Center and Global Regulatory Cooperation Project under the U.S. Chamber of Commerce, and APCO Worldwide, Washington, D.C. Available at http://www.apcoworldwide.com/content/PDFs/Chinas_Drive_for_Indigenous_Innovation.pdf. Last accessed on August 7, 2011.
industries will follow similar trends, in particular, those SEIs related to the advanced computing technologies and products. In overall growth, the added value of SEIs has been estimated to reach U.S. ~$682 billion in 2015 and U.S. ~$1.8 trillion in 2020, with projected annual growth rates of 24.1 percent between 2011 and 2015 and 21.3 percent between 2016 and 2020.32
3.2.3 Transition Toward Economic Outcomes-driven S&T Programs
Unlike most previous government-led science and technology (S&T) programs, where the Ministry of Science and Technology (MOST) was the only or biggest stakeholder, SEI program development efforts have primarily been led by the National Development and Reform Commission and Ministry of Industry and Information Technology (MIIT), two superministries with strong economic missions in China’s bureaucracy.33 In contrast to MOST, these agencies are expected to increase industry participation in the program (as opposed to primary participation by universities and research institutes). As such, it is likely that future S&T programs will require strong economic components and targets and cannot be assessed by publications and patents alone. The fact that economic instead of science agencies are deeply involved in the SEI Program reveals a long-standing issue—the separation between research and the economy—that China has tried to solve when it started to reform its S&T management system in the mid-1980s. Regardless, the successful transformation of the Chinese economy will depend upon the successful coordination among different government ministries, each with a unique mission. Achieving this will be difficult, if not impossible.
In the last several decades, Taiwan has experienced incredible economic growth—particularly in its IT industry, which now accounts for 70 percent of Taiwan’s total manufacturing R&D.34 To date, a defining characteristic of Taiwan’s IT industry has been its deep integration into diverse and global corporate networks of production and innovation. In addition to facilitating the role of Taiwanese firms as fast followers, this network integration also encouraged IT firms to focus their R&D efforts on incremental innovation. Today, there is a growing recognition that Taiwanese firms must now increase and broaden R&D in order to avoid diminishing returns of network integration. Thus, new policies are needed to spur domestic capabilities for low-cost innovation in IT.
While Taiwan’s new innovation strategy is still a “work in progress,” some major policy building blocks, which combine market-led innovation and public policy coordination of multiple layers of private and public innovation stakeholders, are taking shape. Due to its pragmatism and openness to new forms of public policy and private-public partnerships, Taiwan’s innovation policy may in fact shed new light on the opportunities and challenges for strengthening America’s innovation capabilities in advanced computing.
3.3.1 Taiwan’s “Global Factory” Innovation Model35
Early on, Taiwan’s IT industry depended heavily on international markets and access to foreign technology, tools, and ideas to overcome substantial entry barriers to network participation, namely, a lack of domestic market and limited resources and infrastructure. From the beginning, the key to Taiwan’s success has been an early integration into diverse and constantly evolving network arrangements that include both formal corporate and informal knowledge networks. Formal corporate production networks link Taiwanese firms to large global brand leaders (the customers), investors, technology suppliers, and strategic partners through foreign direct investment as well as through venture capital, private equity investment, and contract-based alliances. Equally important are informal global knowledge networks that link Taiwan to more developed overseas innovation systems and knowledge communities, primarily in the United States, through the international circulation of students and knowledge workers.36 Finally, domestic interorganizational linkages with large Taiwanese
32Zhou Zhizue, chief economist on MIIT projections. Available at http://tech.sina.com.cn/it/2011-08-04/07185880063.shtml. Last accessed on November 7, 2011.
33Barry Naughton, 2009, “China’s Emergence from Economic Crisis,” China Leadership Monitor, No. 29. Available at http://media.hoover.org/sites/default/files/documents/CLM29BN.pdf. Last accessed on November 7, 2011.
34Between 2001 and 2006, almost 90 percent of the R&D investment of Taiwan’s private sector was concentrated on two sectors, electronics components (56 percent) and computers and electronic and optoelectronic products (32 percent). See http://eng.stat.gov.tw/ct.asp?xItem=6503&CtNode=2202&mp=5.
35Sections 2.3.1- 2.3.4 rely heavily on Dieter Ernst, “Upgrading through innovation in a small network economy: insights from Taiwan’s IT industry”, Economics of Innovation and New Technology, June, 2010 Volume 19, No. 4pp. 295-324. Last accessed August 21, 2012.
36Between 1987 and 2003, this small island has been the fifth largest nation of origin of international students in the United States. (Guo, 2005: 142).
business groups complement these international linkages.37
Progressive integration into these diverse production, knowledge, and innovation networks has enabled Taiwanese firms to combine the speed and flexibility of smaller firms with the advantages of scale and scope that normally only large firms can reap, as well as to tap into the world’s leading markets, especially in the United States. Network participation has also multiplied conduits for knowledge transfers to Taiwanese IT firms, broadening their scope for learning and capability development. This, in turn, has created new opportunities, pressures, and incentives for Taiwanese IT firms to upgrade their technological and management capabilities and the skill levels of workers.
Today, Taiwan has established itself as an important “global high-tech factory” for PC-related products, handsets, wireless equipment, integrated circuits, and flat panel displays.38 For global IT industry leaders, Taiwanese firms have become preferred OEM and ODM (original design manufacturing) suppliers.39 In addition, Taiwanese firms have made considerable progress in product development, especially in electronic design. Beginning in the 1980s, Taiwan’s leading PC firms established R&D laboratories in Silicon Valley to gain early access to the product and technology road maps of the global industry leaders and to improve their product-development capabilities. By the mid-1980s, Taiwan’s semiconductor firms became involved in board-level and application-specific integrated circuit design,40 giving rise to a broad portfolio of design implementation capabilities. This enabled Taiwanese semiconductor firms to compete on the speed, cost, flexibility, and quality of providing these services.
However, over the last decade, the globalization of S&T and of the economy has placed pressures on Taiwan’s IT industry. Investments necessary for radical innovations that can compete with global technology leaders are beyond the reach of many Taiwanese IT companies. Even TSMC (Taiwan Semiconductor Manufacturing Company), the world’s leading IC foundry, has had to stretch its resources to the limit to sustain its leadership position. In addition, Taiwanese IT firms are establishing low-cost supply bases in China and Southeast Asia to reduce production costs. To expand their position as network suppliers, Taiwanese firms are also moving beyond the provision of manufacturing services, and developing integrated service packages that include logistics and product development. However, the downturn in the global electronics industry since late 2000 has exposed several challenges for Taiwan’s innovation model.
3.3.2 Negative Effects of Network Integration
Taiwan’s focus on the provision of OEM and ODM services has severely constrained the capacity of Taiwanese firms to invest in “upgrading through low-cost innovation” strategies.41 This problem is exacerbated by relentless pressure from global brand marketers to reduce cost and time-to-market for commodity-type products with low profit margins that are apt to penetrate mass markets. As a result, Taiwanese firms are stuck in a “commodity price trap,” with insufficient profit margins to support investment in new R&D, intellectual property creation and branding. Furthermore, as specialized OEM and ODM suppliers, Taiwanese firms typically concentrate on incremental innovations within existing product architectures that are predefined by global brand leaders charging hefty patent-licensing fees. This structure constrains their capacity to develop new products and to shape technology road maps and standards.
Since 2003, many Taiwanese handset makers have attempted to increase profits by increasing their branded handset sales relative to their OEM or ODM business. However, with the possible exception of ASUS and HTC,42 most of these attempts seem to have failed, causing Taiwanese handset makers to switch back to the OEM-ODM model. The most spectacular failure has
37D. Ernst, 2001, “Small Firms Competing in Globalized High Tech Industries: The Co-Evolution of Domestic and International Knowledge Linkages in Taiwan’s Computer Industry,” in P. Guerrieri, S. Lammarino, and C. Pietrobelli (eds.), The Global Challenge to Industrial Districts: Small and Medium-Sized Enterprises in Italy and Taiwan, Edward Elgar, Aldershot, U.K.
38D. Ernst, Upgrading through innovation in a small network economy: insights from Taiwan’s IT industry, Economics of Innovation and New Technology, Vol. 19, No. 4, June 2010, pp. 295–324.
39An OEM contract refers to arrangements between a brand-name company (the customer) and the contractor (the supplier), where the customer provides detailed technical blueprints and most of the components to allow the contractors to produce according to specifications. In ODM arrangements, the contractor is responsible for design and most of the component procurement, with the brand-name company retaining exclusive control over marketing.
40D. Ernst and D. O’Connor, 1992, Competing in the Electronics Industry. The Experience of Newly Industrialising Economies, Development Centre Studies, OECD, Paris, 303 pp.
41D. Ernst, Upgrading through innovation in a small network economy: insights from Taiwan’s IT industry, Economics of Innovation and New Technology, Vol. 19, No. 4, June 2010, pp. 295–324.
42HTC has successfully developed own-brand touch-screen smartphones, initially based on Microsoft’s Windows Mobile operating system, but now also on Google’s open-source Android platform.
been the attempt by the BenQ Group (a spin-off of the Acer Group) to accelerate its global branding strategy by acquiring the mobile handset business of Siemens and its intellectual property.43
3.3.3 Constraints to Developing Indigenous Intellectual Property
While Taiwan’s patent filings at the U.S. Patent and Trademark Office (USPTO) have grown rapidly (in “all patents” per million of its population and in “utility patents”44), its patent counts are highly concentrated, both in terms of products (technology classes) and patent holders (assignees). In 2010 the largest number of Taiwan’s U.S. patents45 were in semiconductor manufacturing, and these patents were dominated by TSMC (Taiwan’s third-largest patent filer, with 405 patents), followed by MediaTek (no. 5), Macronix (no. 6), United Microelectronics Corporation (no. 7), and Via Technologies (no. 9, with 108 patents). Hon Hai (Taiwan’s largest patent filer, with 572 patents), the world’s largest maker of electronic components, has pursued an aggressive strategy to file protective patents (especially for its connector technology), primarily against China.46 In addition, Taiwan’s patent quality remains low (in patent citation, “science linkages,” and technological capabilities),47 and its most influential patents are highly concentrated with TSMC.
Taiwan’s IC design industry provides a telling example of the substantial challenges of developing indigenous intellectual property. As specialized suppliers to global semiconductor and system companies, Taiwanese chip design firms have limited resources and incentives to close the technology gap relative to industry leaders—and as a result, they are typically not active at the leading edge of process technology and IC complexity.48 In addition, Taiwanese design houses have not been able to develop in-house complete solution packages. For instance, in the important cellular chipset market, only one Taiwanese design house (MediaTek) offers a complete cellular chipset solution. All other Taiwanese companies competing in this market, such as Sunplus and Airoha, have focused on specific building blocks and niche markets. Given the rapid change and unpredictability of these markets, such a focused approach is a high-risk strategy.
3.3.4 Hollowing-out Through Offshoring to China
To retain its position as OEM and ODM suppliers to global brand marketers, Taiwan has established low-cost supply bases—and more recently, R&D centers—in China and Southeast Asia. The increasingly common practice of Taiwanese IT manufacturers receiving orders in Taiwan and shipping manufactured goods from China has given rise to “a new cross-Strait division of labor along the lines of pilot run vs. mass production.”; however, this offshore outsourcing now imposes severe pressures on Taiwan’s IT industry, as reflected by a declining domestic value-added ratio that is much lower than for the United States and Japan. 49,50
Another important concern is the continuing relocation of wafer fabrication capacity. Although China’s current wafer fabrication capacity represents about 9.4 percent of worldwide wafer fabrication capability,51 as of May 2011, 28 new wafer fabrication facilities were under construction in Greater China. Once these begin production, it is estimated that China and
43Less than 1 year after the acquisition, the German subsidiary, BenQ MobileGmbH & Co OHG, was closed amid continuing huge losses at the subsidiary. BenQ’s share of the Taiwan handset market now languishes around 8 percent. BenQ now outsources handset production to Taiwanese contract manufacturers.
44A utility patent protects any new invention or functional improvements on existing inventions (such as going from light-emitting diode [LED] technology to organic LED technology), while a design patent protects the ornamental design, configuration, improved decorative appearance, or shape of an invention. China’s utility model patents protect any new technical solution relating to the shape and/or structure of a product, which is fit for practical use. Utility patents, which offer the same protection (albeit for a shorter time span) as invention patents, are quicker and cheaper to obtain, because they only receive a preliminary examination rather than the full substantive examination of an invention application.
45See http://www.uspto.gov/web/offices/ac/ido/oeip/taf/asgstc/twx_ror.htm. Last accessed on January 12, 2012.
46Hon Hai has been expanding its USPTO patent portfolio, more than doubling its USPTO patent filings between 2006 and 2010. Since 1995, 61 percent of Hon Hai’s patents were filed in China, against less than 18 percent in the United States.
47Xin-Wu Lin, 2005, An Analysis of Taiwan’s Technological Innovation – On the Basis of USPTO Patent Data Analysis, slide presentation, Taiwan Institute of Economic Research, Taipei, July 27. For instance, Taiwan’s patents are less “original” and have less “impact” than Korea’s, that is, they are less frequently cited within a technology class and in other technology classes. As for science linkages, Taiwan’s patents, even for semiconductors, are less frequently cited in scientific journals than Korea’s patents.
49This hollowing-out effect, and the resultant job displacements, may have been reduced by the growth of Taiwanese exports to Asia (especially China) of increasingly sophisticated production equipment.
50S.-H. Chen, M.-C. Liu, and K.-H. Lin, 2005, “Industrial Development Models and Economic Outputs: A Reflection on the ‘High Tech, High Value-Added’ Proposition”, manuscript, Chung-Hua Institution for Economic Research, Taipei: p. 25.
51SEMI Wafer Fab Watch, May 2010.
Taiwan together will have a 29 percent share of total worldwide wafer fabrication capacity.
As Taiwanese offshoring extends beyond manufacturing into product development, the competitive advantages previously afforded by Taiwan’s high-tech cluster (i.e., combination of flexibility, low cost, and timely service) have begun to erode. For example, as production of computer, communications, and consumer products moves to China, Taiwan’s IC design houses are forced to follow suit to sustain close interaction with their customers. Once in China, Taiwanese design houses face intense competition from lower-cost Chinese competitors, and they lose their most fundamental competitive advantage: access to a pool of highly trained and experienced lower-cost engineers and managers from diverse sources. Even worse, Chinese IC design firms can now draw on Chinese returnees who have studied and worked in the United States, as well as recruit former employees of Taiwanese companies to train China’s growing pool of local engineering graduates.
3.3.5 Government Policies to Support Low-Cost and Fast Innovation
There is a growing consensus in Taiwan that an exclusive focus on hardware manufacturing is no longer sufficient to guarantee sustainable growth. Taiwan’s new innovation strategies now seek to build on its capacity for low-cost and fast manufacturing by complementing its contract manufacturing and component production excellence with knowledge-intensive support services and a capacity to provide “integrated solutions.” In addition, Taiwan has a long-term objective to strengthen its software capabilities, especially for the design of complex system software and for cloud-computing applications.52 To implement this strategy, Taiwan’s innovation policies seek to strengthen further the linkages and interactions among industry, academia, and public and private R&D organizations.
A defining characteristic of Taiwan’s innovation policy is its openness to foreign strategic advice and knowledge sharing, distinguishing it from Japan, Korea, and China53 with their much more closed systems of innovation policy.
In addition to providing aggressive tax incentives,54 Taiwan’s innovation policy seeks to strengthen the lead role of the private sector by generating new public-private partnerships and by coordinating their interactions.55 In particular, government initiatives, such as Taiwan’s Technology Development Programs, Hsinchu Science Park, and Industrial Technology Research Institute (ITRI), are intended to foster industrial upgrading through low-cost and fast innovation. Today, Hsinchu Science Park is the world’s leading cluster for semiconductor manufacturing. ITRI also continues to play a significant role in Taiwan’s IT and semiconductor industries. ITRI’s recent Cloud Computing Center for Mobile Application (CCCMA) seeks to promote Internet-based, on-demand computing (cloud computing) as a catalyst for strengthening Taiwan’s software capabilities, building on Taiwan’s strengths in lower-cost hardware, such as memory, chipsets, server, and storage network equipment.
3.3.6 U.S.-Taiwan-China Linkages
Since its inception, Taiwan’s IT industry has greatly benefited from its deep integration with America’s innovation system, especially Silicon Valley. As a byproduct, the United States and Taiwan have developed a strong mutual dependence on each other’s IT and semiconductor industries. U.S. IT companies remain the most important buyers of Taiwanese ODM and OEM services, and Taiwan’s silicon foundries are a critical supplier of process technology as well as manufacturing and design services to U.S. fabless design companies. In addition, Taiwan exploits a first-tier supplier advantage due to the establishment of leading U.S. R&D centers in Taiwan and to the acceleration of its “upgrading through innovation” strategy.
However, these relationships have been complicated by the emergence of China as both a partner and competitor with Taiwan. In the last decade, China has become not only the most important production site for Taiwan’s IT companies, but also a major growth market. Not only are Taiwan’s foundries, IC design houses, and ODM suppliers well placed to exploit China’s rapid-
52Interview with Dr. Tzi-cker Chiueh, General Director, ITRI-CCCMA, April 25, 2011. As is typical for Taiwan’s leading innovation actors, Dr. Chiueh’s education and employment history shows strong links with the United States. See also Ministry of Economic Affairs (MOEA), 2011, Taiwan’s ICT industry development and outlook, as reported in DigiTimes, August 29.
53While China has, to some degree, followed the Taiwanese low-cost and fast innovation model, the Chinese model differs in that it has not leveraged domestic and, to a lesser extent, global innovation networks (see Run of the Red Queen: Government, Innovation, Globalization, and Economic Growth in China by D. Breznitz and M. Murphree, 2001).
54Taiwan’s Statute for Industrial Innovation has lowered the business tax from 25 percent to 17 percent in 2010 (which compares to China’s 25 percent rate, Korea’s 22 percent rate, and Singapore’s 17 percent rate).
55H.-S. Chu, 2007, “The Taiwanese Model: Cooperation and Growth,” in C. W. Wessner, ed., Innovation Policies for the 21st Century. Report of a Symposium, The National Academies Press, Washington, D.C., p. 120.
demand growth for IT products and services, but Taiwan’s government is convinced that China is gradually becoming a regional technology leader. This reliance has resulted in new initiatives for cross-strait cooperation in industrial standards, for broader bilateral economic cooperation, especially through the Economic Cooperation Framework Agreement (ECFA),56 and for deregulation of Chinese investment in Taiwan. On the other hand, continuous penetration of the Chinese market will require that Taiwanese firms also redeploy new product development and research to China. By providing critical inputs (through training, technology transfer, and joint product development) to Chinese firms, Taiwan accelerates China’s ability to catch up.
As Taiwan’s IT industry becomes increasingly integrated with China’s economy and its innovation system, it is unclear how and to what degree Taiwan will strike a balance between cooperation with China and cooperation with the United States. If the sheer weight of China forces Taiwanese firms to give priority to their links with China, how will this affect America’s access to the semiconductor global value chain? It is too early for a conclusive answer to these questions. So far, however, Taiwan’s economic diplomacy related to the IT industry remains closely aligned with the U.S. position.57
If Taiwan is to survive intensifying technology-based global competition, it must move beyond its traditional “global factory” innovation model, which will require quick access to radical innovations, especially in generic technologies. While Taiwan has significant policy initiatives in each of the above areas,58 the risk of failure remains high, implying that an exclusive focus on technology leadership strategies is unlikely to support a broad-based upgrading through innovation strategy. These risks explain why Taiwan’s new innovation strategy emphasizes low-cost and fast innovation through domestic and global innovation networks. Recent policies suggest that China is following suit with Taiwan’s innovation model and will focus in the future on low-cost mass adoption of new technologies and innovation.
Countries with emerging economies must rely primarily on foreign sources of knowledge as the main vehicle of learning and capability formation. International linkages are needed to pave the way for an effective exploitation of latecomer advantages. Empirical research has shown that, as a developing country progresses in its industrial transformation, its reliance on international technology sourcing and knowledge linkages substantially increases.60 The Korean innovation system in the electronics industry is emblematic for a heavy reliance on international linkages, combined with the development of complementary domestic linkages.61
Early on, as a part of its innovation strategy, the Korean “government encouraged some of the leading chaebol62 to focus on learning and knowledge accumulation through a variety of links with foreign equipment and component suppliers, technology licensing partners, OEM clients, and minority joint-venture partners.”63 In addition, much of Korea’s success lay in its firms’ abilities to develop the knowledge and
56ECFA is a special free-trade agreement between Taiwan and China, which was concluded in September 2010.
57For instance, during the November 2011 Asia-Pacific Economic Cooperation meeting in Honolulu, Taiwan supported U.S. proposals to extend the Information Technology Agreement and to establish an Environmental Goods and Services Program.
58On SoC design, the government has initiated a National SoC Research Program. On nanotechnology R&D, the government has committed substantial funds, while ITRI and the National Science Council have signed an agreement to conduct joint research with the National Research Council of Canada. And Sha et al. (“ITRI’s Role in Developing the Access Network Industry in Taiwan” in H. S. Rowen, M. G. Hancock, and W. F. Miller (eds.), 2008, Greater China’s Quest for Innovation, Shorenstein APARC, Stanford, CA) describe ITRI’s role in the industry-level upgrading of Taiwan’s access network industry.
59This section heavily relies on Dieter Ernst, “Global Production Networks and the Changing Geography of Innovation Systems: Implications for developing Countries.” East-West Center Working Papers, No. 9, November 2000. Available at http://scholarspace.manoa.hawaii.edu/bitstream/handle/10125/6074/ECONwp009.pdf?sequence=1. Last accessed August 21, 2012.
60For instance, Ernst, Ganiatsos, and Mytelka (eds.), 1998, Technological Capabilities and Export Success - Lessons from East Asia, Routledge, London and New York.
61D. Ernst, 2000, “Catching-Up and Post-Crisis Industrial Upgrading. Searching for New Sources of Growth in Korea’s Electronics Industry,” in F. Deyo, R. Doner, and E Hershberg (eds.), Economic Governance and Flexible Production in East Asia, Rowman and Littlefield Publishers. Taiwan provides another, albeit very different, approach to the development of network integration services through international linkages.
62Chaebol refers to South Korean business conglomerates that are global multinationals owning numerous international enterprises.
63D. Ernst, 2000, “Catching-Up and Post-Crisis Industrial Upgrading. Searching for New Sources of Growth in Korea’s Electronics Industry,” in F. Deyo, R. Doner, and E Hershberg (eds.), Economic Governance and Flexible Production in East Asia, Rowman and Littlefield Publishers. Taiwan provides another, albeit very different, approach to the development of network integration services through international linkages.
skills necessary to monitor, unpackage, absorb, and upgrade foreign technology. Equally important was a capacity to mobilize the substantial funds for paying technology licensing fees and for importing best-practice production equipment and leading-edge components. Most Korean producers arguably would have hesitated to pursue such high-cost, high-risk strategies had they not been induced to do so by a variety of selective policy interventions by the Korean state. By providing critical externalities such as information, training, maintenance and other support services, and finance, the Korean government has fostered the growth of firms large enough to overcome high entry barriers.
It is this coevolution of international and domestic knowledge linkages that explains Korea’s extraordinary success. It has enabled Korean firms to reverse the sequence of technological capability formation. Rather than proceeding from innovation to investment to production, they focused on the ability to operate production facilities according to competitive cost and quality standards.64
Through reverse engineering and other forms of copying and imitating foreign technology, as well as integrating into the increasingly complex global production networks of American, Japanese, and some European global flagship corporations, Korean firms were able to avoid the huge cost burdens and risks involved in R&D and in developing international distribution channels.
For Korea, international linkages provided an important initial catalyst for the development of a sufficiently broad portfolio of domestic capabilities that are needed to reap potential benefits of latecomer advantages.
3.5.1 The Seventh Framework Program for Research and Technological Development (FP7)
The 2007–2013 Seventh Framework Program (FP7) for research and technological development is the European Union’s main instrument for funding research in Europe.65 With a total budget of €53.2 billion, the FP7 aims to increase Europe’s growth, competitiveness, and employment through initiatives and existing programs that finance grants to research actors all over Europe, usually through cofinancing research, technological development, and demonstration projects. However, access to funding is restricted to organizations based in the European Union. This restrictive approach to international cooperation in science and technology is further emphasized in the European Commission’s (EC) 2010 policy document Innovation Union to “ensure that leading academics, researchers and innovators reside and work in Europe and to attract a sufficient number of highly skilled third country nationals to stay in Europe.”66
The 2012 FP7 Work Program is the EC’s largest funding package (about €7 billion) under the FP7 so far and will provide funding to EU-based universities, research organizations, and industries, with special attention given to small and medium enterprises. In addition, it is expected to create around 174,000 jobs in the short term and nearly 450,000 jobs and €80 billion growth in gross domestic product (GDP) over 15 years. Since the initiation of the FP7 Program, investment in industrial R&D by the European Union’s top 1,000 companies has grown by ~10 percent. Between 2010 and 2011, industrial R&D in the European Union grew by almost 6 percent, compared with higher growth reported for the United States (~10 percent), Taiwan (~18 percent), Korea (~21 percent), Hong Kong (~29 percent), and China (~30 percent), and lower reported growth in Japan (-10 percent). These industrial R&D investments can also be broken down by industry classification—of specific interest to this study are the R&D contributions from the telecommunications, software, computer science, semiconductors, and electronics industry. These sectors make up ~23 percent of the European Union’s67 and of Japan’s industrial R&D investments, compared with ~35 percent for Hong Kong, ~39 percent for China, ~41 percent for the United States, ~41 percent for India,68 ~70 percent for Korea, ~77 percent for Singapore, and a staggering ~94 percent for Taiwan.69,70
In R&D intensity, however, the European Union continues to lag behind Japan and the United States. At
66EC, 2010, Europe 2020 Flagship Initiative Innovation Union. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. Available at http://ec.europa.eu/research/innovation-union/pdf/innovation-union-communication_en.pdf, page 27. Last accessed on January 7, 2012.
67Among the European Union member states, Germany, France, and Finland made the largest R&D investments in the telecommunications, software, computer science, semiconductors, and electronics industry sectors (16.1 percent, 22.4 percent, and 82.8 percent, respectively).
68India’s R&D investment sectors consist primarily software and computer services.
70See http://iri.jrc.ec.europa.eu/research/scoreboard_2008.htm. Last accessed on January 7, 2012.
1.6 percent, the European Union’s 2010 share of R&D expenditure in GDP trails both Japan and the United States by a considerable margin, with 3.3 percent and 2.7 percent shares, respectively.71 Among the member states, Germany dominates—at 2.5 percent, its share of R&D expenditures in GDP is much larger than the European Union share. More importantly, however, Germany was deemed to have the highest propensity to innovate.72,73 Hence, it is important to emphasize that national innovation policies differ quite substantially across Europe, both in their overall strategic vision, and in their effectiveness.
3.5.2. Toward an Integrated EU-wide Innovation Strategy
Germany’s move toward an integrated innovation strategy74 is emblematic for a growing trend within the European Union to adopt a much more centralized approach to innovation. In 2000 the European Union established the European Research Area (ERA) to promote a “single innovation market.” One of its main objectives was to optimize and open European, national, and regional research programs to support the best research throughout Europe and to coordinate these programs to address major challenges together.75
In December 2008 the Competitiveness Council adopted a 2020 ERA vision, which seeks to increase the Europe-wide mobility of innovation capabilities by promoting the free circulation of researchers, knowledge, and technology. In 2010 the EC developed an integrated innovation strategy entitled “Innovation Union” to tackle three main challenges for EU innovation policy: (1) underinvestment in knowledge foundation (e.g., the United States and Japan are out-investing Europe and China is rapidly catching up); (2) unsatisfactory framework conditions, ranging from poor access to finance, high costs of intellectual property rights (IPR) to slow standardization and ineffective use of public procurement; and (3) too much fragmentation and costly duplication.76
3.5.3 The European Union’s Key Enabling Technologies (KET) program
An interesting attempt to operationalize Europe’s integrated innovation strategy is the European Union’s Key Enabling Technologies (KET) Program.77 The EC’s six KETs—nanotechnology, micro- and nanoelectronics, advanced materials, photonics, industrial biotechnology, and advanced manufacturing systems78—were selected based on their economic potential, their value-adding and enabling role, and their technology and capital intensity with R&D and initial investment costs. KETs are defined as “knowledge and capital-intensive technologies associated with high research and development (R&D) intensity, rapid and integrated innovation cycles, high capital expenditure and highly-skilled employment.”79 KETs are also embedded in advanced products and they underpin innovation chains.
Advanced computing products, such as multicore processors and parallel software developments, are examples of technologies that are consistent with the KET definition. Like other KETs, advanced computing technologies provide potential first-mover advantages, and enable the owner of relevant intellectual property rights to create new lead markets as new technologies replace old technologies with few or no other players. One of the key goals of the European Union’s KET Program is to reduce the deeply ingrained barriers to industrial innovation. In other words: Why are breakthrough ideas, inventions, and discoveries (that
71Battelle, 2010, “Global R&D Funding Forecast” in R&D Magazine, December 2009. Available at http://www.rdmag.com/uploadedFiles/RD/Featured_Articles/2009/12/GFF2010_ads_small.pdf. Last accessed on August 11, 2012.
72An innovation, here, is defined as a new or significantly improved product (good or service) introduced to the market or the introduction within an enterprise of a new or significantly improved process.
73Eurostat, 2010 Yearbook, p. 606. Available at http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-CD-10220/EN/KS-CD-10-220-EN.PDF. Last accessed on August 15, 2012.
76EC, 2010, Europe 2020 Flagship Initiative Innovation Union. Communication from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions, Commission Communication (COM(2010)546). Available at http://ec.europa.eu/research/innovation-union/pdf/innovation-union-communication_en.pdf, page 27. Last accessed on January 7, 2012.
77EC, 2011, High-Level Expert Group on Key Enabling Technologies. Final Report, June. Available at http://ec.europa.eu/enterprise/sectors/ict/files/kets/hlg_report_final_en.pdf. Last accessed on August 15, 2012.
78EC, 2009, Preparing for our future: Developing a common strategy for key enabling technologies in the EU, Commission Communication (COM(2009)512). Available at http://ec.europa.eu/enterprise/sectors/ict/files/communication_key_enabling_technologies_en.pdf. Last accessed on August 15, 2012.
79EC, 2010, Current situation of key enabling technologies in Europe, Commission Staff Working Document (SEC(2009)1257). Available at http://ec.europa.eu/enterprise/sectors/ict/files/staff_working_document_sec512_key_enabling_technologies_en.pdf. Last accessed on August 15, 2012.
were developed with public R&D funds) not transformed into commercially successful innovations within reasonably short time frames?
3.5.4 Policy Options
To overcome the above deeply entrenched innovation barriers, the European Union’s KET Program proposes a broad range of coordinated support policies that cover the following stages of the “innovation chain,” from the transformation of fundamental research into globally competitive technologies, through product development to make innovative and cost-effective product development and prototyping, to globally competitive manufacturing.
Specifically, the EC KET Program identifies the following five priority areas for Europe’s evolving EU-wide innovation strategy: (1) sustain a critical mass in knowledge and funding through effective use of economies of scale and scope; (2) increase market focus of R&D projects; (3) invest in large-scale demonstrators and pilot test facilities; (4) provide post-R&D commercialization support; and (5) practice trade diplomacy, that is, reduce unfair subsidies and protect domestic companies from unfair trade practices.80 This last policy priority is of particular concern from a U.S. perspective. In fact, the European Union’s KET Program culminates in a fairly “techno-nationalist” notion of IPR protection and states that “the EU should clearly promote an ‘in Europe first’ IP policy” and that proposals require clear IP plans for “first exploitation of IP” and rules that “favour the EU exploitation of the results of projects.”81
The European Union has experienced a fundamental change in its innovation policy from government-centered national strategies to attempts to combine market-led innovation and public policy coordination across Europe. While government initiatives, such as the KET Program, attempted to bridge the perennial gaps that stymie Europe’s industrial innovation ecosystem, significant challenges remain. To a large degree, however, this transformation is still a work in progress, as European IT innovation and commercialization continue to lag.
In addition, there are signs that Europe’s fiscal crisis and increasingly severe austerity policies might slow down Europe’s move towards greater openness and internationalization of its innovation system.
The diversity of economic and IT innovation policies across the United States, China, Taiwan, and Europe reflect their differing cultures and history, economic status and technical capabilities. The U.S. approach rests on government support for basic academic research and a vibrant capital market and private enterprise ecosystem for product innovation. The other countries and regions blend elements of private enterprise and central planning. Each is unique and not directly transferrable to another region. Nevertheless, there are some general principles that can be gleaned from this survey of policies, coupled with technical insights regarding semiconductor device fabrication, chip architecture, and software.
Some of the largest computing companies in the United States have internal multidimensional technological capabilities in chip design, process development, wafer manufacturing, and software and have demonstrated success tapping into foreign talent pools and markets. However, IT talent, capabilities, and facilities are increasingly distributed globally. Although research prowess is correlated with industry success, information flows globally via many sources. The lesson of basic research, both in industry and academia, has been that the discoverers are not always those who convert the ideas into economically successful products. Oftentimes, the likelihood that an idea can be successfully commercialized and implemented depends on a nation’s or region’s innovation policies and entrepreneurial climate.
Second, the cost of semiconductor fabrication facilities is rising exponentially, placing their construction beyond the economic reach of small- and mid-sized companies. Only the largest multinational companies and nation-states can fund their construction and operation. This suggests that the United States must be mindful of its global dependence on fabrication supply chains and that it develop realistic models that balance the need for the latest process technology versus multiaxis innovation and that combine reliability and resilience, programmability, and functionality. Although financial investment in fabrication facilities by a small number of U.S. companies, primarily by Intel, provides some domestic sourcing, most of the chips contained in devices sold in the United States are fabricated offshore. IBM does produce some chips in the United States, both for U.S. defense needs and its own products, but the volume is relatively small.
Third, there is no assurance that historical U.S. dominance in computing will transfer to new and emerging domains. The need for architectural and software innovation to deliver new features and greater
80EC, 2011 KET, p. 33.
81EC, 2011 KET, p. 37.
performance via parallelism creates opportunities for new ecosystems to emerge and evolve. With licensable components and global access to fabrication facilities, it is possible for this innovation to occur almost anywhere. In addition to performance as measured by computing speed (clock speed, bandwidth, interconnect, and so on), it may be that other measures—such as reliability and resilience, programmability, security, and efficiency—become equally, or potentially more, important. For example, efforts to improve programmability and efficiency of base processors might yield significant improvements in software quality, software development times, and (ultimately) application performance.
Fourth, global policy makers see information technology in general and consumer computing in particular as major economic forces to be harnessed for local and regional benefit. They are investing in the future, hoping to position their region for success. Which of the myriad approaches being pursued will be most successful is difficult to predict.
Today is an inflection point, when the virtuous cycle of faster sequential processors has broken down and when new devices and services are emerging to reshape the computing landscape. An intense global competition for IT hegemony is under way. No company, country, or region will reap all of the economic benefits, as the global value chain is too intertwined for that. However, there will be economic winners and losers, just as there always are whenever technology shifts occur. U.S. policy makers would be wise to think carefully and deeply about the shifts under way and their implications for economic competitiveness and national security.