From YalePressWiki

[edit] Chapter 3: The Development of the IT Industry in Taiwan: Public Research Institutions as Growth Impetus?

If you want to build something new where nothing was before, go to the U.S. If you want to build something that would last for twenty years, go to Japan. If you want to work successfully under always-changing conditions and changing regulatory systems, go to Taiwan.
—A founder and manager of one of Taiwan’s first IT companies

Even within the world’s most distinguished group of successful emerging economies, the East Asian Newly Industrialized Countries (NICs), Taiwan has one of the most inspiring stories. Taiwan is the only society in the region that has, in many critical aspects, closed the gap in innovational activities with the leading Western industrial nations and with Japan.^[1] Taiwan has also developed a vibrant industrial system of indigenous new small and medium-sized enterprises, a system that is not dominated by a few huge conglomerates or subsidiaries of foreign MNCs (Wu 2001, 2005). Taiwanese companies have become a prominent force in the IT industry’s global production network. During the 1990s Taiwanese companies pioneered the model of the pure integrated circuit (IC) silicon chips fabrication (pureplay) foundry in semiconductors, and more than half of the world’s silicon chips are now being fabricated in Taiwan. Designing and manufacturing products for the worldleading MNCs, Taiwanese IT companies in 2001 manufactured 70 percent of the world’s motherboards, 55 percent of laptops, 56 percent of LCD monitors, 51 percent of color-display tube monitors, and 36 percent of digital still cameras.^[2] In addition, with 225 companies and NT$147.8 billion in sales, Taiwan’s IC design subsector is the second-largest in the world after the United States’. The total sales of the Taiwanese electrical and electronics industry in 2002 reached almost $88 billion USD, and on top of that the semiconductor industry added another $21.4 billion.^[3]

How did this transformation happen? In Taiwan the state itself, acting as the technology-creating agent, has spurred the growth of the IT sector; the Taiwanese state acquired the necessary technological information and skills for high-technology industrial development policy by embedding the technology-creating agents within its structure. The main policy vehicle behind this achievement was the establishment of two public research institutions responsible not only for making the decisions on what R&D to conduct but also for implementing and executing them.

Nonetheless, the Taiwanese approach to S&T industrial policy was not without its limitations, in particular the stagnation of the software sector and the limited new product–innovation capabilities developed by the private industry. In addition, under the Kuomintang (KMT) as the governing party, the state purposely built a financial system with a severe scarcity of patient capital, in stark rebuttal to Gerschenkron and the developmental state models following him, specifically Taiwan’s neighbors, South Korea and Japan (Amsden 1989, 2001, Gerschenkron 1962, Johnson 1995, Woo-Cumings 1999). Taiwanese companies’ average equity-to-debt ratio has been as low as in the United States and, at times, even lower (Cheng 1990, 1993, Fields 1997, Gold 1986). This helped the Taiwanese economy to remain relatively unscathed from the East Asian financial crisis (Park 2000). However, it limits the ability of Taiwanese companies to compete in many IT manufacturing sectors (Fuller et al. 2003). Even in comparison with Israel and Ireland, the state has taken limited financial commitments (Cheng 1990, 1993, Pack 1993, Park 2000).

Taiwan has also followed a distinct policy with regard to MNCs and the shaping of the relationships between its local IT industry and global markets. Early in its industrial efforts, its policy goal was to become a supplier of components, systems, and design-and-manufacturing services to the global IT production network. Thus the story of the spectacular and unique growth trajectory of the Taiwanese IT hardware industry, and the less than spectacular growth of the country’s software industry, stems not from massive investment of patient capital but from a distinctive institutional system of industrial innovation— a system built around public research institutions–led R&D science-and-technology (S&T) industrial policy, and a policy goal of attaining a defined position within the IT industry’s global production network.

In this chapter I analyze the general growth of the Taiwanese IT industry and the Taiwanese S&T industrial policies. I pay special attention to the successful political processes of co-evolution between the Industrial Technology Research Institution (ITRI) and the private hardware industry, and the unsuccessful process of co-evolution managed by Institute for Information Industry (III) and the software industry. I also analyze the state’s strategy in acquiring the necessary technological skills and information and its effects, and the state’s attempts to control and plan the technological development path of the IT industry. Lastly, I inquire into the changing roles of the state in MNCs’ activities in Taiwan, and their relationships with Taiwanese companies, within and outside Taiwan.

My conclusions are, first, that the success of the hardware sector and the relative failure of the software sector stem from the different politics of stateindustry co-evolution in the two sectors. In hardware, ITRI evolved from being the initiator and leader of the industry to a position of a supporting actor, in the process instigating the move of the Taiwanese industry to leadership positions in semiconductor manufacturing and design, as well as in design and manufacturing of desktop, laptop, and other electronic systems and components. In software, from the beginning, III evolved to a position of a competitor to the private industry, doing as much to contend with it as to support it. III never managed to take a contributing leadership position in any software technology development project in the same way that ITRI did in hardware. Thus the Taiwanese software industry that developed did so in spite of III, not because of it. These differences originate, I argue, from the different positioning of the two institutions by the Taiwanese state, as well as from some inherent differences in the industrial sectors for whose development they were nominally responsible.

My second conclusion is that the institutional system of Taiwan, a system with strong division of labor between industry and public research institutions, limits the R&D capabilities of private firms by motivating these firms to concentrate more on development than on research. This system developed from the particular decisions of the state on how to acquire its own technological skills and knowledge. In this system the public research institutions decide which R&D projects should be pursued, conduct the lion’s share of R&D, and only then diffuse the results to the private industry; accordingly, core R&D capabilities development is concentrated almost solely within the public sector. Thus in Taiwan, specifically because of this institutional system, the industry has focused on, and successfully developed, world-leading capabilities in second-generation, process, and manufacturing innovation, but not in new-product development.

My third conclusion is that the decision of the state to focus its attention on building a local supplier network for MNCs, coupled with the way it constructed its capital markets, paved a development path for an IT industry that concentrates on several complementary activities of the global IT industry’s network: components development and design, together with manufacturing and OEM and ODM services. This development path correlates well with the IT innovation system, which has been developed around the private-public division of labor, with its strength in second-generation innovation and integrated design capabilities. However, it remains to be seen whether Taiwan can, or should, move to original new-product R&D activities in the future, the main declared goal of government policy in the past decade.

[edit] Historical Background

Looking at the successful Taiwanese IT industry at the beginning of the third millennium, one can easily forget that in the 1950s Taiwan was a poor agriculture-based island with almost no industrial or technological infrastructure, ruled by what was for 85 percent of its population a foreign force, the defeated mainland China KMT. Taiwan’s emergence as a democratic society has been equally remarkable since its politics started from foreign domination, including the KMT’s organized killing of many of Taiwan’s native intellectual elite (Gold 1986).

Much of the literature that attempts to explain Taiwanese economic growth depicts Taiwan as a “poster child” neodevelopment state. In almost all accounts, Taiwan is described as a society in which the state takes the lead in new economic activities—always, however, with the aim of spurring private entrepreneurs and always through interactions with private industry (Amsden 2001, Amsden and Chu 2003, Evans 1995, Fields 1997, Fuller 2002, Hong 1997, Hung 2000, Mathews and Cho 2000, Wade 1990, Wang 1995, WB 1993). In most accounts, Taiwan’s industrial and innovation systems come through as a humming machine of industrial development, with a smooth and well-functioning division of labor between government and industry. It is this division of labor that is considered responsible for Taiwan’s leading role in the global information technology industry. Indeed, from the point of view of the development of the IT hardware industry, Taiwan seems to be an exemplary case of a state managing to handle the complex, changing, and sometimes contradictory roles our theoretical approach argues that a state needs to handle in order to spur the growth of innovation-based industry.

Three recent accounts of Taiwan’s high-technology industrialization exemplify this point by reaching an almost complete agreement on the ways in which the state and industry in Taiwan have co-evolved in developing the IT hardware industry since the early 1970s (Amsden and Chu 2003, Hong 1997, Mathews and Cho 2000). While each author has a different vocabulary, each describes a similar two-phase process. First, the state’s research agencies acquire a technology from abroad, absorb it and improve it, and then spin off private companies to spur the industry. Second, after the private industry has emerged, the industry and state settle on a new division of labor in which the state’s role is to locate and absorb new technologies, infuse the industrial system with them, and assist private firms with their own advanced R&D projects. However, the actual sociopolitical-economic history of these developments is far from smooth.

[edit] The 1960s: The World’s Semiconductor Industry Going Global

Taiwan’s modern IT industry dates to the 1960s.^[4] At that time, the new semiconductor industry had evolved in the United States to become the first to have truly globally fragmented production. During the 1960s new developments made it feasible to geographically disconnect the heavily low-skilled, labor-intensive final assembly stages from the rest of the production processes. At the time Taiwan opted to follow an MNC–foreign direct investment (FDI) industrial policy similar to Ireland’s, and the state marketed Taiwan as a business-friendly place with an unlimited supply of stable, reliable, hardworking labor and a growing market (Klintworth 1995). In these years, Taiwan had many of the same attractions for Western companies that mainland China now seems to possess.

With the help of the United States Agency for International Development (USAID), the Taiwanese state was actively targeting foreign MNCs, following the same tactics that were to become the hallmark of the Irish Industrial Development Agency (IDA) in later years. Extensive background work was performed on individual executives to create a feeling of personal connection. Executives of U.S. firms coming to Taiwan were treated like honored guests; they were ushered to personal meetings with ministers who were thoroughly briefed on the visitors’ personal backgrounds and sometimes even managed to display symbolic pictures, emphasizing a common past, in their offices (Wade 1990).

As part of those early efforts at FDI attraction, Taiwan also established export-processing zones (duty-free manufacturing areas), again a striking similarity to Ireland’s first attempt at FDI attraction to the duty-free industrial region around Shannon International Airport (Hwang 1991). The year 1961 marks the beginning of the transformation of global IT into an industry ruled by global fragmented production networks. In that year, Fairchild of the United States and Philips of Holland established their first semiconductor final-assembly factories in Asia, Fairchild in Hong Kong and Philips in Taiwan.

In 1964, after aggressive overtures by the Taiwanese government, General Instrument opened its first facility in Taiwan. In 1965 what the Taiwanese claim to be the world’s first export-processing duty-free zone was opened near Kaohsiung, and between 1964 and 1966 twenty-four U.S. firms joined Philips and GI in opening manufacturing facilities in Taiwan.^[5] These efforts also mark the beginning of the rise to prominence of Taiwan’s two most influential S&T industrial policy leaders, the technologists-politicians Drs. Y. S. Sun and K. T. Li (Cheng 1990, Hong 1997, Mathews and Cho 2000, Wade 1990). In 1965, by creating Taiwan’s first data-processing and software company— the China Data Processing Center—the state also pioneered what would later become a pillar of its IT industrial policy, the establishment and subsequent privatization of public firms in new technological areas.

The FDI-based IT industrial policy proved, at first, to be very successful. Exports of electronics boomed at 58 percent annual growth rates between 1966 and 1971. However, relatively quickly it became apparent that the U.S. semiconductor companies had no intention of locating any higher-skilled operations in Taiwan. The sustainability of this first success in semiconductor and advance electronics relied on relative low wages (Amsden and Chu 2003, Wade 1990). With wages rapidly rising as a result of the early economic success, it became apparent that unless something was done, this success in semiconductors would be temporary.

In the same period, and even earlier, the electronics-manufacturing industry in Taiwan was rapidly growing. Here, the Taiwanese government policy sharply diverged from that of Ireland. The Taiwanese state, taking a more nationalistic approach toward ownership, started to demand that a large and growing percentage of the components used for final assembly in Taiwan by MNC subsidiaries and joint ventures be locally manufactured.^[6] The policy proved to be successful in breaking the enclave positions of MNCs. This was especially true in the case of the Japanese MNCs, which aimed to sell to the local market and hence were more susceptible to the new local-components requirements. By 1971, 37 percent of all components used by the electronics industry in Taiwan were manufactured locally. Moreover, the first originalequipment manufacturing (OEM) relationships between Taiwanese companies and foreign MNCs started in these sectors, becoming prominent first in the manufacturing of color televisions (Amsden and Chu 2003).

This policy had far-reaching impacts on the future of the Taiwanese IT industry and S&T industrial policy. The KMT’s early political aim was to prevent the emergence of competing centers of power. This goal led to the creation in Taiwan of a financial system that specifically limited the use of large investment of long-term capital and the accumulation of debt. That in turn led to a particular vision of the future of the Taiwanese IT industry. Accordingly, the first goal of the Taiwanese S&T industrial policy was to spur the growth of a local IT industry that supplies many of the necessary components, especially the higher-value ones, to the global IT production network. The second goal was to develop a local IT industry that provides manufacturing (and later design) services of increasing sophistication to the world’s leading MNCs. It has never been the goal of the Taiwanese state to create big vertically integrated national champions in the same way that other Asian NICs, such as South Korea, did. Nor was it ever the goal to create cutting-edge innovation-based firms similarly to those of Israel or the United States. Of our three cases, Taiwan was by far the most persistent in its efforts to convince MNC manufacturing subsidiaries to embed themselves more deeply within the Taiwanese industry, and in urging them to transfer more and more responsibilities over a growing span of activities to local suppliers.

For these reasons, by the time the economic and political crises of the 1970s hit Taiwan, its electronic industry was already geared toward a specific relationship with, and embeddedness within, the global IT industry production network. Moreover, the Taiwanese innovation system and industrial policy were already centered around components manufacturing, thus on manufacturing and process innovation, not on the generation of R&D-based new products.

[edit] The Crisis of the 1970s and the Crystallization of the Public Research Institutions–Based Industrial Policy: The Creation of the Semiconductor Industry

In the early 1970s the global economy in general and Taiwan’s economy in particular went through several waves of severe economic and political crises. In 1971 the Bretton Woods system was dismantled; in 1973 the first oil shock resulted in a steep rise in the price of oil, and many of the advanced industrial economies started to suffer from stagflation. Taiwan’s exports, then mainly textile, footwear, and low-level electronics, started to face competition from less-developed countries and protectionists’ measures in their target markets. In addition, with the reemergence of China on the global political scene and the derecognition of Taiwan in the United Nations in 1971, the country had started a long period of formal political isolation. It was under those conditions that Taiwan reformulated its S&T industrial policy. Nonetheless, for years many politicians and bureaucrats did not view the IT industry as a primary source of growth. As we shall see, the overall budget of ITRI’s semiconductor technology–development efforts has been minuscule in comparison with the amounts dedicated to other industries by the Taiwanese state and with the investment of other states in their semiconductor industries.

By 1973 some Taiwanese-owned subcontracting firms for assembly and packaging of semiconductors appeared; however, there were still no direct forward or backward linkages, and the semiconductor industry was still an MNC enclave utilizing relative low labor costs and focusing on low-skill final assembly stages. In January 1973 Yun-Hsuan Sun, then the minister of economic affairs and later the premier, oversaw the establishment of ITRI near Hsinchu, a strategic location close to Taiwan’s two leading engineering universities, National Chiaotung University and National Chinghua University. ITRI was created through the merger of three existing governmental labs. Using his considerable political power, Sun managed to transfer the responsibility for electronics R&D from the Ministry of Telecommunication laboratory to the newly formed ITRI. However, as was the case initially with Israel’s OCS, ITRI was not at the time considered to be a major policy initiative, and its overall financing was negligible from the point of view of the state’s overall industrial policies.^[7]

The idea behind ITRI’s creation was to form one main national lab, which would be responsible for the upgrading of Taiwan’s industrial technology through technology transfer, development, and diffusion. The objective was that the government would take it upon itself to solve the risky and more challenging part of the R&D process and would then diffuse the results to the industry, which would concentrate on final development and manufacturing. Thus the state had opted to bring the R&D-creating agents within its own structure, developing a particular division of labor between the industry and the state. This division of labor, which after a few successes became prominent in the innovation system of Taiwan’s semiconductor and electronics industry, has had significant influence on the kind of capabilities private firms developed, and on the kind of business models and business strategies that gained paradigmatic hold in Taiwan.

In August 1974 Sun met his friend Dr. Wen-Yuan Pan, a Chinese-American engineer working at RCA’s David Sarnoff Laboratories in Princeton, New Jersey. The two agreed on a plan to formulate an S&T industrial policy geared toward the creation of a semiconductor industry in Taiwan. Pan subsequently established a group of mostly Chinese-American engineers working for leading American semiconductor companies in the United States. The group regularly convened at Princeton as the Technical Advisory Committee. The committee submitted specific recommendations for the establishment of a specialist lab in ITRI that would act as the focal launching point for the industry. In September 1974 the Electronic Research Service Organization (ERSO) was established in ITRI, with its main first goal being the development of technological capabilities to spur the growth of a semiconductor industry.

ERSO started to look for sources of IC fabrication technologies. However, it did not manage to find any suitable partners until, through Pan’s influence, in 1976 RCA agreed to transfer its obsolete technology to ITRI. RCA had decided to get out of the semiconductor industry and saw this as an opportunity to earn royalties from its abandoned seven-micron complementary metal oxide semiconductor (CMOS) fabrication technology. This technology was already far behind the world limit of two microns. A group of forty engineers, many of whom later became the leaders of the semiconductor industry in Taiwan, spent almost a year in RCA’s facilities in the United States, and ERSO built its first IC fabrication plant with RCA’s guidance. In 1977–1978 the first trial wafers were produced, and the Taiwanese team started to test its own experimental designs. In 1979 the ERSO team advanced to such a degree that it had better yields than RCA’s and started to sell small quantities of chips to supplement its financial resources.

At the same time, the Taiwanese state was establishing the physical and business infrastructure needed for the semiconductor industry. It is here that Kuo-Ting Li’s leadership over Taiwanese S&T industrial policy became crucial. Li was minister of finance after serving as the minister for economic affairs. In 1978 Li organized a special conference on Taiwan’s economic and scientific future. The result of this well-attended event was the document “The Science and Technology Development Program,” which was adapted by the cabinet and called for the creation of (a) a special permanent advisory body for science and technology, the Science and Technology Advisory Group (STAG), which would be chaired by Li and report directly to Premier Sun, and (b) an infrastructure focused on the needs of science-based advanced industries.

The latter project was championed by the president of National Chinghua University, S. S. Hsu, who urged the creation of a science-based industrial park next to Taiwan’s three prominent engineering institutions in Hsinchu to duplicate what he perceived to have happened around Stanford University in California (Hong 1997).^[8] The park, named Hsinchu Science-Based Industrial Park, was put under the jurisdiction of the National Science Council (NSC) and launched in 1980.^[9] Hsinchu Park became an extremely important factor in Taiwan’s subsequent success in the semiconductor industry. Hsinchu Park has been accepting only firms with significant R&D operations, for which, apart from providing specialized infrastructure, it has been granting special subsidies, such as tax holidays, duty exemption on equipment importation and commodity exports, low-interest loans, and matching R&D funds.

However, it was not private industry that established the first leading firm of the Taiwanese semiconductor industry, as well as many of the subsequent ones. In the late 1970s, facing what seemed to be large initial investment with high uncertainty, no private entrepreneur, industrial group, or investor was willing to invest capital in commercializing the technologies developed in ERSO. Under these conditions both Sun and Li decided against the establishment of a fully state-owned company and chose to create a joint private-state venture. In 1978 the head of ERSO, together with one of the forty engineers who went to RCA, proposed a plan for the establishment of a private-state company. The ministry of economic affairs (MoEA) accepted their proposal and invited some of Taiwan’s biggest companies as investors.^[10] At first, none of the companies invited agreed to join. MoEA then used its direct influence and organized a coalition of local companies that finally agreed to invest in the project, and between them, 51 percent of the shares of the new entity, United Microelectronics Corporation (UMC), were distributed. The total sum invested stood between $14 million and $20 million USD.^[11]

Apart from receiving all of its technical staff and technology from ERSO, UMC was also granted technical assistance from, and the use of, ERSO’s fabrication plant, a source for many subsequent conflicts of interest. UMC quickly engaged in another series of fund-raising and built its own first fully owned fabrication plant in the newly established Hsinchu Park in 1982.

[edit] Diversifying and Deepening the Technological Base of the Semiconductor Industry: ERSO's Strategy of Parallel Capabilities Development

ERSO’s preeminence over the development trajectory of the Taiwanese semiconductor industry did not end with the spin-off of UMC. On the contrary, the successful launch of UMC encouraged subsequent formal and less-formal spin-offs, as teams of engineers left to establish their own companies. Following the establishment of UMC, ERSO itself slightly changed its focus. At the urging of American advisers, ERSO made a few important decisions. The first was to stay clear from what would become an extremely capital-intensive market niche: memory chips, specifically DRAM.^[12] The second was to base the Taiwanese semiconductor industry around capabilities that would allow it to competitively and quickly design and manufacture custom-tailored (application- specific integrated circuit—ASIC) chips, and not just around specific products. ERSO argued that by developing this technological capability Taiwan would, first, strategically differentiate itself from Korea and Japan, which had embarked on a strategy of high-volume rapid-generational-upgrading hardware such as memory chips, which necessitate extremely deep-pocketed companies of the kind Taiwan did not possess. ERSO further argued that this strategy would enable Taiwan to focus on the consumer industry, where Taiwanese electronic manufacturing companies were already globally successful and realized substantial revenues using relatively simple technology. ERSO rightly contended that by developing these capabilities, Taiwan’s semiconductor industry could innovate across the whole spectrum of the IT industry (Wade 1990).

The Taiwanese semiconductor industry developed capabilities that both (a) strengthened the IT hardware OEM and ODM industry and (b) allowed it to excel in the market niches opened to it by the existence of Taiwanese OEM and ODM firms. In other words, the decision by ERSO to prompt the development of strong ASIC capabilities in the semiconductor industry positioned the industry perfectly to take advantage of the opportunity structure offered by the successful IT design and manufacturing subcontractors industry, and strengthened the competitiveness of that industry in turn. As I shall show, this in turn has reinforced Taiwan’s embeddedness in the global IT production network as a supplier and innovator of components and design-and-manufacturing services and processes.

As part of its strategic decision to focus on ASIC technologies and capabilities, after UMC’s spin-off ERSO pursued a dual-capabilities development strategy. The strategy focused on developing both ASIC capabilities and the fabrication technologies needed to manufacture them. ERSO managed to move its own fabrication capabilities to 4.5 microns in 1981 (still far behind the world’s technological edge at the time at 2 microns), and to acquire and develop complementary technologies that would enable Taiwan to excel in ASICs. In the early 1980s ERSO expanded its capabilities in design, testing, and several other stages of IC production, including masking. It was in this early stage that many teams of ERSO employees started to leave ITRI to open IC design houses that utilized ERSO’s and UMC’s fabrication facilities, one of the first being Syntek in 1982 (Chang and Tsai 2002). Hence, even before the establishment of the pureplay foundry model by another ERSO spinoff— Taiwan Semiconductor Manufacturing Corporation (TSMC)—more than a few design houses were operating in Hsinchu Park, and within a year after the establishment of TSMC their number was estimated at forty.

[edit] Transforming the Global Semiconductor Industry: Establishing the Pureplay Foundry Model in Taiwan

The pureplay foundry business model, pioneered by ERSO and spun off as TSMC in 1986, developed as part of ERSO’s efforts to infuse the Taiwanese industry with very large–scale integration (VLSI) capabilities. ERSO’s foreign advisers, who did not view the progress to 4.5-micron technology as fast enough to close the gap with the world’s technological leaders, urged these efforts on. In 1981 STAG, chaired by Li, argued that Taiwan should achieve VLSI capability of 1 micron or better, the technological frontier at the time.

A political battle immediately ensued, as this view was attacked from almost all quarters in Taiwan’s economic development bureaucracy, including the Ministry of Finance, the Council for Economic Planning and Development, and a large percentage of ERSO’s personnel. However, with the backing of now-Premier Sun and Li, a decision was made to go forward with a VLSI 1-micron technology development project simultaneously with ERSO’s other activities. In 1983, $72.5 million USD was allocated to a five-year program on various semiconductor projects, a larger amount than provided before but still insignificant in both international terms and in comparison with other Taiwanese industrial policies. The goal was to reach 1-micron capabilities by 1988. Once again, ITRI was leading the private market.

The decision on the implementation of the VLSI research project was a continuation of the earlier strategy of a division of labor in R&D between public institutions and private companies, with the public institutions commencing most of the research and the results diffused to private companies that focused mostly on development and design. After a battle with UMC, which argued that it should be responsible for the development of the VLSI technology, ERSO was again the chosen agency, in part to prevent an overconcentration of power and capabilities in one private company (Fuller et al. 2003, Hong 1997, Mathews and Cho 2000, Meany 1994).

This time ERSO decided to follow a joint-venture technology-development strategy and did not pursue technology-transfer contracts. The chosen party was Vitelic, a small Silicon Valley firm founded by Chinese-American engineers in 1983. The R&D phase was successful, with a DRAM chip developed by June 1986. However, at the time, neither the government nor the private sector in Taiwan was willing to supply the necessary manufacturing facility, and as a result Vitelic sold its technologies to various Korean and Japanese companies (Meany 1994).

This awkward situation politically strengthened those who argued for more direct state intervention in the building of VLSI fabrication facilities in Taiwan (Hong 1997). In addition, by the end of 1986, with many of the new design houses in Hsinchu Park opting to follow Vitelic and selling their technologies to Korean and Japanese firms, it became obvious that there was a wider demand for VLSI fabrication facilities.

In the same time that ERSO developed VLSI technologies, UMC developed its own similar technology, utilizing both in-house development and joint-venture agreements. This technological competition was straining even more the conflict of interest between ERSO and UMC, especially as Morris Chang, the head of ITRI, was also serving as the chairman of UMC. Chang also epitomizes another important trend in Taiwan’s rise to prominence in the semiconductor industry, the returning immigration wave of graduate level– trained engineers from the United States.^[13]

Taiwan has one of the highest numbers of engineering and science students who immigrated to the United States to get advanced engineering and science degrees. For most of the period until the early 1980s, these Taiwanese expatriates did not see any opportunity to employ their skills profitably in Taiwan, and most of them stayed on in the United States. However, with the creation of ITRI, and even more so with the growing success of the semiconductor industry, a large percentage came back or established cross-country businesses. This so-called reverse brain drain has been one of the most important factors in the rapid growth of the semiconductor industry (Saxenian and Hsu 2001).

However, while the influence of the returnees has been important, we should remember that this pullback started and strengthened only after many of the émigrés saw an economic opportunity in Taiwan, not vice versa. Moreover, while returnees established many companies, locals founded most of the key companies.^[14] In addition, while the sizable return from Silicon Valley is celebrated, the influence of returning immigrants from other parts of the United States, notably the large contingent coming back after long stints in Bell Labs and IBM, has been at least as important, a fact that our interviewees repeatedly noted during our research trips to Taiwan (Amsden and Chu 2003, Saxenian and Hsu 2001).

In 1986, after some internal debate, a decision was made to move forward with the privatization of ERSO’s VLSI efforts and the construction of a VLSI fabrication facility employing an innovative business model—the pureplay foundry. This decision marks a turning point not only for the Taiwanese semiconductor industry but also for the global industry. As at UMC, the decision was to commercialize and privatize the VLSI project by creating a joint public-private company, TSMC. Again the Taiwanese state faced a similar investment landscape as with UMC. Although by now the success of UMC and the growth of the industry was proof that the semiconductor industry in Taiwan could be profitable, the private market was still unwilling to commit to a leading investment.

This lack of will by Taiwanese investors ironically proved to be a boon for TSMC, as Philips, the Dutch MNC, agreed to provide 27.5 percent of TSMC’s initial capital and became its leading private shareholder, with another 48.3 percent of the investment coming from the China Development Corporation, an investment arm of the ruling KMT party (Fuller 2002, Fuller et al. 2003). Apart from securing some business from Philips, having the MNC as its main shareholder gave the new venture legitimacy; even more important was a full patent swap agreement, which was critical in the beginning in protecting TSMC from legal actions.^[15]

The pureplay model had a major revolutionary influence on the global IT production network, on Taiwan’s embeddedness within it, and on the reshaping of Taiwan’s semiconductor industry. The pureplay foundry model is one of the best examples of an organizational innovation that enabled new levels of fragmentation of the semiconductor industry’s production network—an innovation spurring a rapid process of advances in production-stage specialization and product-stage economies of scope and scale, consequently in stage-specific innovation. Before the establishment of TSMC and the pureplay foundry, the semiconductor industry consisted mainly of companies that utilized the business model of integrated device manufacturers (IDM). The IDM model calls for the creation of large vertically integrated firms that perform both design and fabrication in house. The leading semiconductor firms were all IDMs that built their own dedicated fabrication plants. The industry also consisted of design houses that were smaller and marginal and that needed to secure their fabrication capacity from IDMs with no standards for information transfer, and with the added risk of sharing their IP with potential competitors. Moreover, the process was lengthy and cumbersome, putting the design houses at a distinct disadvantage.

The pureplay foundry model calls for the creation of companies whose sole business is fabrication. Pureplay foundries receive codified designs from the design houses and fabricate their chips for them. Thus they enable stage specialization in both design and fabrication. In 1986 the viability of the pureplay foundry was still unclear, as the technology for full codification of the designs was not yet developed. However, in the particular case of Taiwan—(a) the industry’s growth limited by the political choice of a financial system that restricted the use of long-term financing and debt; (b) the already growing number of design houses in Hsinchu Park; (c) the political pressure arising from the lack of VLSI fabrication capability that brought many of these design houses to sell outright the IP to foreign firms; and (d) the political constraints that the existence and lobbying of UMC put on ERSO—the pureplay foundry model seems to be worth the commercial and technological uncertainties that still exist. An additional advantage of the model was that a pureplay foundry could also learn from its customers. Many of TSMC’s early customers passed on their technologies to TSMC so that it could fabricate their chips. They were willing to do so only because TSMC’s pureplay foundry model allayed their fears of its becoming their competitor. Consequently, TSMC has been using customers’ feedback to refine and expand its fabrication method (Fuller et al. 2003).

For a few years, TSMC’s growth was slow. Nevertheless, even then its effect on the Taiwanese semiconductor industry was immense. Its services enabled a growing number of fabless companies to profitably commence operations. This unleashed the commercial potential of the Taiwanese industry’s ASIC specialization and skills. By 1994 there were already sixty-five design houses in Taiwan (ITRI 2003, various years). In that year the design-codifying and transfer technology was fully developed, and since 1997 the pureplay foundries have been fabricating the lion’s share of chips in Taiwan, with many of the IDMs, including UMC, converting to the foundry model or being bought by pureplay foundries in need of more fabrication facilities.

Pureplay became the prominent production model in the global semiconductor industry, completely transforming it in several ways. First, the pureplay model enables groups of engineers to quickly bring innovative designs to market with only limited financial resources, fostering even more rapid innovation cycles in the semiconductor industry. Second, the lower cost also enables groups of engineers to profitably utilize the business strategy of offering cheaper chips in already developed niche markets using second-generation technologies. Third, the pureplay model transformed the industry by creating virtuous cycles of internal innovation and growing efficiency that make it even more profitable, vis-à-vis the IDM model. Today, thanks to their head start, the Taiwanese pureplay foundries TSMC and UMC are the global market leaders, with sales in 2004 of just less than $8 billion USD and just more than $4 billion, respectively. TSMC and UMC are the main reasons behind the growth of the Taiwanese semiconductor industry to its current status as the world’s fourth-largest IC producer.

Apart from UMC and TSMC, ERSO was also responsible for a few more of the leading Taiwanese semiconductor firms before it was closed and the Computer and Communication Lab (CCL) became ITRI’s sole semiconductor technology division. The Taiwan Mask Corporation (TMC) was spun off ITRI in 1989, culminating the successful development of the masking technologies project at ERSO, which had begun with technology transfer from the United States in 1977. TMC successfully grew in the years since, and in 2000 it bought Innova, Taiwan’s second masking company.^[16]

Two other important spin-offs from ERSO were Vanguard and Winbond. Vanguard was the official spin-off of ERSO’s submicron project that culminated in an industrial-size fabrication plant in Hsinchu Park. It was auctioned in 1995, with TSMC becoming its main shareholder. While Winbond is not an official ERSO spin-off, its history makes its connection to ERSO clear. In 1987, after the launch of TSMC, ERSO was still in possession of its first pilot fabrication plant and its personnel. The plant manager, Dr. Ding-Yuan Yang, engineered an MBO, secured with financing from the Walsin Lihwa Corporation, launching it as Winbond. Unlike other Taiwanese companies, Winbond has been following the OBM model and has stayed as an IDM company, moving into both memory and logic. In 2004 Winbond’s sales were slightly less than $1 billion USD.

[edit] From OEM to ODM: The Technological Upgrading of the Electronics Suppliers, and ITRI's Role as a Supporting Actor

Taiwan’s ODM-OEM firms are as important to Taiwan’s success in the IT industry as are its semiconductor firms, if not more so. Taiwan’s OEM-ODM firms are successful in all sectors of the IT industry, employing many different business models. The range of strategies moves from more focused OEMODM companies—such as Quanta or Inventec, which design and manufacture most of the world’s laptops, and Asustek, which does the same in motherboards—to more flexible OEM-ODM firms such as Sampo, BenQ (formerly part of the Acer group), Tatung, and Mitac, which operate across a wider array of products.

The existence in Taiwan of local world-leading OEM-ODM companies also propelled the growth of the semiconductor industry in a particular trajectory. The existence of ODMs has given the Taiwanese IC design industry a competitive advantage and a unique opportunity structure focusing on second-generation innovation. These opportunity structures are not available to Israeli, Irish, and American IC design companies. The OEM-ODMs, in turn, have been strengthened by the success of the local IC design industry, which caters specifically to their needs and helps them to improve their quality while lowering their costs.

ITRI and the Taiwanese state have been as important to the success of this sector as they were to the semiconductor industry. However, while ITRI was the first and, for a long time, the only actor in the field in the semiconductor industry, in the case of the OEM-ODM electronics, ITRI’s main goal has been to technologically upgrade existing firms, some of which, like Tatung (1918), have been around since early in the twentieth century. This sector gives us an excellent example of how the Taiwanese state has pursued the same goal —developing a local IT industry that supplies many of the necessary components and manufacturing services to the global IT production network, applying the same division of labor in R&D, while using a different mechanism as the state-led research consortia. In addition, the case of OEM-ODM enables us to follow another angle of the co-evolution of industry-state relationships.

As has been vividly argued by John Mathews, the main goal of the Taiwanese R&D consortia has comprised technological learning, upgrading, and catch-up as part of attempts at industry creation (Mathews 2002). This is a very different objective from the strategy of Israel, described in Chapter 2, of constructing a special program to sponsor collaborative R&D efforts in an attempt to enable smaller Israeli firms to successfully compete with MNCs on the basis of their own developed cutting-edge technologies.^[17] Like Israel, the Taiwanese state recognized the handicaps of its IT industry—in this case, that the smaller size of Taiwan’s manufacturing-oriented companies, with their lack of sophisticated in-house R&D capabilities, hinders the ability of the industry to upgrade and compete in global markets. The solution was to assemble a series of ITRI-led research consortia. From the start of this process, the same division of labor in R&D that had been used in the case of semiconductors was applied, with ITRI conducting the R&D and the private companies concentrating on manufacturing and final product development. Even in more recent research consortia, in which private companies have assumed leadership roles, this division of labor has been maintained.

The first few consortia, and some of the most significant ones, were developed as part of ITRI’s multiclient projects in the early 1980s. These projects assisted in the creation of a PC development-and-manufacturing industry in Taiwan. The first such project started after Apple clamped down on the thriving Taiwanese Apple II cloning industry in late 1982. The industry, then a highly fragmented one with no prior experience with IP protection laws, came to MoEA and ITRI for help.^[18] ERSO already possessed some experience working with Intel processors, and ITRI advised the industry to concentrate on the new IBM Intel-based PC instead of Apple (Noble 1998). The real activity started after February 1983, when Acer approached ERSO with a request to develop a PC.

After various halts and requests from other computer-manufacturing companies, ERSO devised the project as its first multiclient project, with the title of MCP-1. A decision was made to limit the number of participants to nine. By the end of 1983 Taiwanese companies had started manufacturing PC clones, hoping to ship them to the United States in time for Christmas sales. However, IBM counterattacked, claiming that ITRI’s basic input-output system (BIOS) was infringing on its IP.^[19] ERSO engineers rewrote the code, but by the time IBM agreed that the new version was legal, it was already May 1984. By then, most of the project’s participants had either sold their PCs in less IP-strict markets, such as Europe and Asia, or followed Acer and used a different BIOS (Noble 1998).

Although this project was deemed less than satisfactory, as soon as IBM announced its new PC AT system in August 1984, ERSO announced its intention to launch a new multiclient project. This time the participants were limited to three, including Tatung, which used this project as an entry into the computer business. The prototype was transferred by ERSO to the companies by July 1985, and soon thereafter the companies started to bring their own products to market (Chang et al. 1999, Mathews 2002, Noble 1998). This project was considered a great success and paved the way to a series of related multiclient projects. From this point onward the projects were run by the newly established Computer and Communications Laboratory (CCL) and were more formally viewed as consortia, not as ITRI’s multiple-clients research projects.

By 1990 the Taiwanese industry was firmly established in desktop computers and other related products; nonetheless, its technological capabilities were still far behind what was needed to succeed in the lucrative new laptops market. CCL formally drafted the project as an R&D consortium with the aim of developing a set of key components to become the standard on which the different companies would build and develop their products. The biggest computer-manufacturers trade association in Taiwan was involved and recruited as a joint coordinator. In July 1990 the Taiwan Laptop Consortium, with forty-six subscribing companies, was officially announced with capital of $2 million (USD; Mathews 2002). The project was concluded quickly, and by the end of 1990 Taiwan achieved the status of the world’s leading supplier of laptops. This time CCL not only transferred prototype machines but also constructed extensive training programs, and many of its staff moved to private firms. The main problems of the project were oversubscription and a too tight division of labor. The companies, given an almost complete product that had been developed solely by CCL, were unable to differentiate and further develop their products, leading to a relentless price competition among the Taiwanese companies. These lessons were learned, and throughout the ITRI-led consortia in the 1990s, private companies were given slightly greater R&D responsibilities, or at least slightly more-flexible prototypes. In some of the projects one of the declared aims was the creation of R&D, or at least integrateddesign capabilities, within the participating private companies.

As the Taiwanese industry has grown to become globally important, many of the leading MNCs have happily supported many of ITRI’s consortia. The MNCs have hoped to spur the growth of an industry based on their technology, and find it easier to deal solely with ITRI than separately with every private company. The best examples of such a project was NewPC, which was based around the PowerPC processors developed jointly by IBM, Motorola, and Apple (Dedrick and Kraemer 1998, Mathews 2002).

However, in the key area of developing the R&D capabilities of private companies, the research consortia are still far from successful. A top official of CCL bitterly described the situation:

Technology transfer from CCL is very important in Taiwan. The companies use us to train their people, take our technology, and create some products. After which they go to their customers, show them the products as a proof that they can do OEM, sign some OEM contracts, and just do the same things for a couple of years down the road. Our aim is to help and spur them to routinize R&D and product innovation. However, the end result is that they do not do it, a main reason being that they can more cheaply rely on technology transfer every time they need to upgrade their skills, which allow them to operate as a pure OEM. (interview, May 24, 2001)

The state also has used the research consortia mechanism not only to develop and create new IT products industries but also to upgrade the Taiwanese electronic components industry, following its strategy from the early 1960s of trying to localize the creation of key components.

One of the most illuminating such projects was the CD-ROM project, which evolved from MoEA’s decision in 1992 to view CD-ROM as a strategic product. As a result of this project, by the end of 1999 Taiwan moved from being a supplier of less than 1 percent of the world’s CD-ROMs to supplying more than 50 percent, with almost 49 million units sold (Amsden and Chu 2003, p. 104). This project evolved into a series of consortia and moved on to DVD technology.

From the point of view of our argument about opportunity structure, it is important to note that one key subcomponent that none of the consortia managed to develop was the IC chip. By 1997 the development in the CDROM industry, along with ITRI’s inability to develop the needed IC chips, created an ideal market opportunity for a new IC design company, and MediaTek was spun off of UMC. MediaTek succeeded where ITRI’s consortia failed and quickly developed chips for the CD-ROM industry, breaking the monopoly of Japanese companies in the area. MediaTek’s technologies moved together with the industry’s, first to CD-RW and then to DVD.^[20] In the process, MediaTek became Taiwan’s most successful IC design company and the world’s fourth-largest in terms of sales since 2002.

Nevertheless, even in the IT hardware sector not all of the Taiwanese state’s initiatives have been successful. In parallel with ITRI, the Taiwanese government has sponsored and expanded an even bigger pubic research institute— Chungshan Institute of Science and Technology (CSIST)—Taiwan’s defensetechnology research institute. Apart from funds received through the defense establishment, CSIST was, as late as 1999, the receiver of the second-largest amount of funds, after ITRI, from MoEA’s Department of Industrial Technology (Hsu and Chiang 2001). Little information is available on CSIST, but while ITRI’s efforts resulted in significant commercial success for relatively low investment, those of CSIST, often working on similar technologies with vastly larger resources, did not achieve even one commercial success in the semiconductor industry.^[21] This compares poorly not only with ITRI but also with Israel’s experience. The main difference between CSIST and both ITRI and Israel’s RAFAEL is that CSIST does not maintain close relationships with the private industry and instead prefers to work in isolation and secrecy.

[edit] The Development of the Semiconductor Industry in the 1990s: The Growth of the IC Design Subsector

Of the many subsectors of the global IT hardware industry in the 1990s, the IC design (or fabless) subsector was one of the most innovative and most successful.^[22] By the early 2000s Taiwanese IC design houses were the only non–North American companies to reach the top ten positions in terms of worldwide sales. As a matter of fact, Taiwan and Israel are the only non–North American countries to have any companies at all among the top thirty companies. By 2003 the Taiwanese IC design industry was already the second-largest in the world in terms of sales after that of the United States. The IC design industry was the focus of attention for the Taiwanese state in the 1990s and has been key in the attempts of the Taiwanese state to spur more cutting-edge innovational activity.

As can be seen in table 3.1, between 1991 and 2002 the Taiwanese IC design industry experienced a remarkable growth. From an industry that consisted of 57 companies in 1991, the industry grew to 66 companies employing 2,109 people in 1995, and to 225 companies employing 11,800 in 2002. This rapid growth, with sales of more than $6 billion USD in 2003, made it one of the three biggest IC design industries in the world (ITRI 2003, various years).

Looking at the largest ten companies by sales in 2002–2004 (table 3.2) and the thirty largest fabless companies in 2002–2003 (table 3.3), we see that the industry’s skills are quite developed and flexible, as attested to by the ability of Taiwanese IC design firms to excel in almost every subsector of the semiconductor industry. In addition, it is apparent that these successful companies have built their businesses around the opportunity structures offered by the existence of the large OEM-ODM companies in Taiwan and ITRI’s efforts to develop local manufacturing of key components. For example, MediaTek’s core business has been the PC’s CD and DVD industries; VIA is developing PC chipsets for motherboard and PC manufacturers; ALi’s main business is laptop chipsets; and Sunplus is catering to the consumer markets, especially to toy OEM-ODMs.

Another common trait appears if we analyze the strategy of the leading companies: none of them develops completely new innovative products that are first to the market. Furthermore, even the most innovative firms, those that manage to develop some brand recognition for their products, such as VIA and ALi, rely on developing secondary innovations on the basis of technologies developed by the leading American MNCs. This tends to leave many of them in a weakened position, where changes in technology or in the behavior of leading MNCs can rapidly deteriorate their market position.

The story of VIA, the largest Taiwanese IC design company in terms of sales until 2002, is a case in point. PC chipsets are VIA’s main product line. As long as Intel’s Pentium 3 was the leading PC CPU, VIA gained global market share at Intel’s expense, reaching almost 50 percent, in part because of Intel’s decision to stick with the more expensive Rambus technology. With the move to Pentium 4 technology, Intel attacked VIA on multiple fronts: it did not license VIA its P4 technology, and it took VIA to court over patent infringement allegations in October 2001. As a result of these moves, all first-tier PC manufacturers, such as Asus or Gigabyte, refrained from using VIA’s chipsets, fearing legal action from Intel. VIA countersued Intel for both patent infringement and usage of monopoly status, and the legal battle was fought in five jurisdictions. In April 2003 the two companies reached a settlement that includes a patent swap agreement and a licensing agreement. By that time, however, VIA had lost substantial market share and revenues. Moreover, because it agreed to pay royalties to Intel, it can no longer retain the profit margins it enjoyed before. Although in 2001 VIA was the biggest Taiwanese IC design house, with about $1 billion USD in sales and control of almost 50 percent of global market share, in 2002 VIA was only the second-largest Taiwanese IC design house, with sales down to $729 million and global market share of less than 25 percent. In 2003 VIA’s sales continued to drop, and its sales were only $600 million (AFX 2001, FT 2003a, 2003b, 2003c, Hille 2003, Hung 2002).

Another example of the vulnerability of the Taiwanese IC design industry position within the global production networks is the fact that this legal battle was fought after VIA had already changed its ownership structure to placate Intel’s demand. In order to do so VIA separated itself from its main shareholder, First International Computers (FIC). This restructuring was done because Intel viewed the combination of motherboard producer, FIC, and chipset producer, VIA, as a threat.^[23]

As can be seen from the rapid growth of the design, or fabless, subsector after 1996, the main impetus for growth had been the establishment and rapid growth of the pureplay foundries. Both major foundries, TSMC and UMC, have taken leading roles in the development of the design subsector. Both established exclusive foundry-IC design house relationships with what they term club members, which they also assisted financially. Many of the top IC design houses, such as MediaTek and Novatek, were spun off from UMC when it turned its business model from IDM to pureplay foundry.

[edit] Structure, Development, and Business Models in the Taiwanese IC Design Industry

Taiwanese fabless companies employ one main business model, apart from the few cases in which a Taiwanese company manages to sell its own branded products to final customers. This business model is based on the utilization of the unique advantage that the Taiwanese IT industry’s structure offers: the colocation of both the world’s biggest and most advanced pureplay foundries and of some of the largest OEM-ODM companies.

Most of the Taiwanese IC design houses employ two variants of the same strategy: to supply the Taiwanese (and more recently the Chinese) OEMODMs with the chips they need in order to supply finished products at competitive prices and of sufficient quality to OBM MNCs. The Taiwanese fabless companies rely on the proximity of the world’s biggest pureplay foundries. This geographical proximity enables them to produce cheaply and speedily large quantities of newly designed chips and to inspect and assure quality in almost real time. All the Taiwanese IC design houses whose personnel we interviewed specialize in chips based on second-generation technologies with the aim of lowering costs and increasing reliability; they also do some process innovation. Apart from an unsuccessful attempt by ICreate, a subsidiary of Etron, to develop and market advanced chips in a joint venture with the Israeli company Zoran, no Taiwanese IC design house at which we interviewed tried to develop original products, new technologies, or products that are based on cutting-edge technology. Figure 3.1 provides a graphic representation of the main differences between the market relation and information-and-communication flows of the Taiwanese industry in comparison with the Israeli and U.S. industries.^[24]

The Taiwanese fabless companies’ business model has two variants that are not mutually exclusive. In the first variant, the design houses custom-design chips to clients’ specifications (a local OEM with specific contracts with an OBM), test and quality-assure them, and then manufacture and deliver them using a pureplay foundry for fabrication. In the second variant the IC design house builds “standardized” chips for OEM companies and other customers, manufactures them using a pureplay foundry, and sells them either directly or through distributors. Representatives of the firms whom we interviewed claimed that as the market for standard chips is vastly larger, revenues using the second variant with successful products are much higher. However, because custom-designed chips have the advantage of a secure customer and higher profit margins per fabricated chip, a significant number of companies use both business models.

Figure 3.1. Annual Exports Software and IC Design, Taiwan. Sources: III various years, ITRI 2003, 2005, ITRI various years, MoEA various years.

As the customization-based business model also has lower costs and companies can secure quick revenues, until recently it has been the preferred mode of operation for new companies. The story of Sunplus, consistently one of the top four companies in term of sales since 2000, is illuminating. Sunplus was established by a group of former ERSO engineers who left ITRI to join Silicon Integrated Systems (SiS) and left SiS after its restructuring in 1989. The idea of the founders was to focus on the consumer market, especially toys; thus reliability and price have always been much more important than the latest technology and features. Using their own money to seed the company, the founders immediately sought revenues to supplement their capital and started operations with custom-made ASIC designs. In 1994 Sunplus manufactured its first standardized chip, a microcontroller for toys with LCD monitors, from which it branched out to other multimedia-related chips. Currently, Sunplus uses the customization-based business model for its entire gift and toy business and the standardized-chips business model for the rest.

Most of Sunplus’s revenues now originate from its standard chips; however, as there is fierce competition in the multimedia mass market, a large percentage of profits come from the smaller custom-made-chip division. Sunplus realizes that even though it does not use cutting-edge technologies, in order to excel, it must keep abreast of new technologies and be able to implement them quickly once their prices go down. In order to do so, Sunplus employs a few strategies similar to those of many successful Taiwanese fabless companies. First, it maintains a close relationship with ITRI, especially with its CCL lab, using ITRI as a channel for new technologies. Second, Sunplus has established its own investment arm, which invests in foreign and local start-ups in fields that Sunplus thinks it might need in the future. Third, Sunplus directly licenses technologies.

The Taiwanese IC design houses rely not on in-house-developed cuttingedge IP to give them market advantage but on their ability to deliver moderately sophisticated products more speedily, cheaply, and reliably than their competitors. This is true even for companies that pass through the new government-sponsored small-business innovation research (SBIR) program and/or ITRI’s incubation center. For example, one interviewee told us: “Our game plan is to focus on the China market. The U.S. is far too technologically advanced for us.”^[25] Another interviewee, a serial entrepreneur, the founder of a company that was then still in ITRI’s incubation center, added, “The customers we aim to have need good chips but not state-of-the-art chips. They care more about reliability, mass production, and unit costs than cutting-edge technology. They are not looking for new innovations.” Two other interviewees, from two of the most technologically advanced communication Taiwanese fabless companies, were even more pessimistic about the status of innovative activities in the industry:

There is only one place in Taiwan that develops new technology and cutting edge products: ITRI. I sometimes miss being there. The only reason I left is because I wanted to get rich.
I do not know about any real R&D-based company in Taiwan, apart from ITRI, of course. Everybody is interested in revenues and profits. MediaTek is the biggest model of success. The biggest problem for any R&D-based model to work is the stock options–bonus regulations. These pull all the talent into the big manufacturing companies. (interviews, January 30, 2003)

This is equally true for all the companies that have been established by returning immigrants from the United States, including those that have successfully founded and managed new product–based IC design houses in Silicon Valley. Hence the decision to utilize only the second-generation innovation business model has less to do with skills and capabilities than with the institutional structure of the Taiwanese industry and its links with the financial markets.

In sum, the fabless chip sector evolved around a particular business strategy that has relied on the unique institutional features of the Taiwanese IT industrial system as a whole. The main three features are (a) the existence of large OEM-ODM companies that sell in foreign markets, which creates a large demand for chips based on second-generation technologies; (b) the close proximity of the world’s largest and most advanced pureplay foundries; and (c) an innovation system with a division of labor between ITRI and the private sector that excels in quick technology transfer and second-generation innovation, while continuously infusing the system with the most recent foreign technologies.

Utilizing this strategy, the IC design subsector provides complementary assets to the Taiwanese OEMs and pureplay foundries. First, it supplies the pureplay foundries with a constant stream of orders in a variety of IC designs, helping them to stay profitable and to extend and maintain some of their own technological capacities. Second, it supplies OEM-ODM manufacturers with the chips they need either to profitably offer solutions for Western OBMs or to compete successfully with Western and Japanese OBMs by lowering their cost structure.

The development of the IC design subsector thus enriches the Taiwanese IT industry with positive feedback, strengthening the Taiwanese position as midlevel supplier within the global IT product networks. Each part of the Taiwanese IT hardware industry—OEM-ODMs, pureplay foundries, and IC design companies— strengthens and is in turn strengthened by the existence, outputs, and demands of the others. Despite this virtuous cycle, or maybe because of it, the IC design industry is unable to develop alternative business models for capturing the higher rents that come from original innovative designs.

We should not, however, discount the magnitude of Taiwanese achievements in IC design. Taiwan, a small and until recently technologically backward society, developed the second-largest IC design industry in the world, far surpassing the South Korean, Japanese, and western European industries. The industry is not only growing but also enhancing the overall competitiveness of the Taiwanese hardware industry. Whether or not there is a move to innovative products, the industry’s growth seems secure in the near future.

[edit] The Taiwanese Software Industry

If the story of the hardware IT industry is one of impressive growth and successful policies, the software industry offers a less sanguine tale of industrial policy and growth. Official statistics are inflated, but even they show the one major failure of the Taiwanese software industry: it is mostly a domesticoriented industry. Official figures present a picture of rapid growth in sales from NT$22 billion in 1991 to NT$149 billion in 2001, but exports in 2001 amounted only to NT$16 billion, or a little more than $490 million USD, less than 11 percent of total sales. For comparison, the annual export figure for the whole of the Taiwanese software industry is just slightly higher than the quarterly sales of Amdocs, Israel’s largest software company. These figures are tiny in comparison with those of the hardware sector, or even with those of the IC design sector alone (for example, MediaTek’s sales alone are more than twice as large as even the official export figures of the software industry). Moreover, the annual sale revenues of two of the three largest Taiwanese software product companies, Ulead and Cyberlink, were around $30 million USD in 2003, too small to put them in the league of medium-sized IC design houses in Taiwan, or large software companies in Israel or Ireland. Figure 3.2 compares the rapid growth of exports of the IC design sector, the smallest and latest subsector of the semiconductor industry, with that of the whole software industry, putting this failure in perspective.^[26]

The relative failure of the software sector is even more puzzling if we take into account the amount of high-level software development critically needed by the hardware industry, especially the IC design sector. Thus the success of the hardware sector in the development of what is usually termed embedded software, coupled with the global success of a few Taiwanese software companies, shows that the cause of this failure cannot be explained by lack of software programming skills in Taiwan.^[27]

With this in mind we can understand why a high-ranking official in MoEA mused in one of our interviews, “Look where we are and where other countries like India, Israel, and Ireland are with their software industry. Nobody heard about them thirty years ago. I must admit that we [the Taiwanese] are not doing very well with software” (interview, May 30, 2001).

In striking contrast to the electronics and semiconductor industries and the positive role played by ITRI, the Institution for Information Industry (III) has no stories to tell of ERSO-like successful spin-offs or any other initiatives spurring the creation of an industry. III has had little positive role in the development of the software industry. Indeed, III can be seen as one of the main obstacles to the industry’s development.

K. T. Li provided the impetus behind the establishment of III in 1979. Li hoped that III could play a similar role in software to the one played by ITRI in hardware, and he remained influential behind III for some years after its establishment. Thus III and ITRI were structured in the same way. Moreover, they were established within six years of each other and under the same political leadership. If the structure of the bureaucracy and its policies were the key variable, we would expect a similar industrial outcome. In fact, though, the political process of sectoral industry-state co-evolution moved III and the software industry on a divergent path from that of ITRI and the hardware industry.

Figure 3.2. Market Relations and Communication Environment of the Taiwanese IC Design Houses

Unlike ITRI creation in relation to the development of the hardware industry, III was created after some of the early leading software firms had been established. The private sector, in contrast to the semiconductor industry, was active in the software sector. The agenda and the funding granted to III by the Taiwanese government were vastly different from ITRI’s. First, III was asked to promote the software industry. Second, III was given the task of promoting the use of IT and software throughout Taiwan and was asked to help the government with its own computerization. Finally, III was also asked to generate enough revenue to cover most of its activities. The different agendas proved to be in conflict, with the end result that III transformed itself into one of Taiwan’s biggest IT consultancies and software houses and has been competing directly with private software firms.^[28] Hence the political positioning of III has spurred it to see itself as a “profit center,” while at the same time giving it the regulatory and political backing to disregard the needs and lobbying of the Taiwanese private software industry.

The responses of many software firm executives to our questions about the role of the state in the industry’s development offer the view of III as the biggest hindrance to the industry’s growth. The response of a founder of a financial software company is typical: “The state does not have any positive role; as a matter of fact, III is our greatest competitor. They also compete unfairly. I need to sponsor my R&D from my revenues. They have all their R&D covered by the state” (interview, January 17, 2003). A founder of another software company observed:

In our seventeen years we had direct conflicts with III only once, so I am very lucky. However, I do not think III is good for the industry. They get government money to help the industry and nothing happens. III do not really care about the industry—they just talk and talk but do not do anything for the industry. They do not even properly do the more simple and straightforward task of consulting the government on policy issues, so policy making is all tangled up. Even in the basic task of changing the perception of the software industry in Taiwan they do nothing. Customers think software and software companies should not be paid because they do not see software as a “real” product. (interview, November 6, 2003)

This view of III is shared not only by those we interviewed in private firms and in industry associations but also by officials of other developmental agencies. One official of an industrial R&D agency responded to a question about the role of III in the software industry’s development: “III is a funny organization. It both competes with and tries to assist the local software industry. On a charitable estimate I would say that they compete at least as much as they assist” (interview, January 29, 2003). Another official, a head of a different department, gave a similar response: “I agree that III is a very problematic institution. I think that the problem of III is the confusion or the strange positioning that they are in. They try to compete and help at the same time” (interview, November 4, 2003).

III is well situated to play a role as a facilitator of collective action and organizer of consortia. However, even in this role both researchers and industry leaders criticize III. Noble claims that even in the crucial case of agreeing on a standard for Chinese-writing input-output and internal conversion—a basic need for the software industry’s growth—III’s actions fostered distrust and hampered the industry’s efforts to reach agreement. In the end an international body, on which Taiwan had no influence, settled the issue almost a decade later (Noble 1998, pp. 123–147). This impression of the failure of state efforts is strengthened by our analysis of the software industry. As we shall see, the most successful subsectors are those in which III never intervened.

[edit] Structure, Development, and Business Models in the Taiwanese Software Industry

The structure of the Taiwanese software industry reveals a divide between older companies, which are focused on the development of software applications for big organizations and fiercely compete with III, and a newer cohort of companies developing software technologies, most of which were founded after the success of the hardware IT industry in Hsinchu Park. This division tempts one to contend that there are in fact two Taiwanese software industries. However, the reality is more complex.

As we have seen, it is extremely hard to gather precise sales figures on the software industry in Taiwan. Most of the bigger companies, faced with intense competition from III, have branched out into sales and distribution of hardware and even into completely unrelated fields at times of severe need. Even public game-development companies, such as Summit, earn much of their revenues from bundled sales—the sales of their products on top of sales of DVD and other entertainment electronics. Estimating software sales in Taiwan is also complicated by the fact that PC manufacturers have been for many years some of the biggest software developers.

Nevertheless, the picture that emerges is of an industry in which until a decade ago most of the Taiwanese software companies merged both IT consultancy and bespoke development, as well as the sales of specific software systems (such as automation or finance) that were developed as part of earlier work with customers and then packaged. Most of the PC manufacturers were also successful software producers, especially in the 1980s, when each manufacturer had its own slightly different version of the Chinese input-output system, and when most of the classic PC software packages from Western companies, such as spreadsheets, were unavailable in Chinese.

Because III had captured the big governmental contracts, and the big global IT companies were competing directly on big projects, the industry was unable to develop big software houses specializing in customized development. A remark from a founder of one of the companies that managed to survive these times is illuminating:

Competing with III is like competing with Microsoft. If you compete, you get killed on the spot; if you cooperate, you get less money and you might be killed later, but at least you get some work. However, III always wins more contracts than they can program themselves, so they then subcontract some of them. They will pay me less than what I would get if I could compete on the project in the free market, but I prefer to cooperate with the Microsoft of Taiwan, and not get killed by it. (interview, November 6, 2003)

As a result, most of the industry has evolved around a particular market niche that was neglected or considered too small by III, such as international banking or securities-trading systems. Today many of these companies are realizing that their future lies with mainland China. And here they find that their former relationships as subcontractors of American MNCs in Taiwan are an important asset. For example, a CEO of one of Taiwan’s oldest software companies described his decision to move to China:

The funny thing is that we never wanted to go to China, but our American MNCs partners asked us to come, both HP and Oracle. The best example is HP. They won the whole IT systems project for a big new plant, and they asked us and three other Taiwanese companies to open a China branch and do subcontracting for them. We did that, and before we knew what happened, Oracle asked us to do subcontracting for them when they won a big Chinese state-owned company contract. I think that without the mainland, my company would not have been able to grow at all in the last three years. Only because of China we have a chance to survive. (interview, November 4, 2003)

For the most part, the industry has not been able to develop unique products or strong enough brands to guard their own products when the big software MNCs launched Chinese versions of their products. Thus when Western companies such as Microsoft localized their products for Taiwan and China, Taiwanese products were wiped out of the market.

The one sector that evolved differently is the PC game industry. Unlike other sectors of the software industry, the video game industry had four tremendous advantages that allowed it to thrive in Taiwan. First, III was not at all interested in this sector. Second, video games, unlike many other software products, benefit from certain cultural traits, and Chinese-born gamers were keen on playing “Chinese” games. For example, almost every Taiwanese game company has had a strategy or fantasy–role playing (FRP) game based on the classic tale of the Three Kingdoms. Third, the costs of game development were very low for many years. Most video game companies either started by sponsoring teams of high school students, who developed the games as a hobby on a cost-only basis, or had a mixed development strategy of in-house and semi-independent teams. Fourth, with 7-Eleven and similar popular chains in Taiwan selling local games on the corner of every block, video game companies have had distribution channels that reached every Taiwanese on a daily basis.

It is not surprising that until 2000–2001, video games were deemed by many in both the private and public sectors to be the most successful part of the software industry in Taiwan. All the medium- and large-sized companies, such as Softstar, Summit, Gamania (formerly known as Full Soft), Interserv, and Soft-World, had gone public on the Taiwanese stock exchange in the 2000–2002 period. Since then, the industry has been hard hit, and in 2003 most of the companies that we revisited had either retreated from original game development or were cutting down their development activities to the bare minimum. Apart from the worsening economic situation in Taiwan, the two changes that transformed the business environment of the Taiwanese video game companies were technological: the wide diffusion of CD-readand- write technology and the rapid emergence of online gaming.

These two technological changes have significantly and swiftly lowered the revenues coming from the sales of PC games, with most but not all of the leading companies reporting sales of tens of thousands of copies of their new games at best, instead of hundreds of thousands of copies. These developments coincide with the maturation of the international gaming industry. The quality and technological sophistication needed to make competitive game titles has drastically raised the costs of development per title. In addition, the Korean gaming industry, for many years the poor and unsophisticated cousin of the Taiwanese industry, was the first to develop online FRP games, winning market share and financial backing and overtaking the quality and technological sophistication of the Taiwanese industry at a critical moment. At the end of 2003 all but one of the extremely successful massive-multiusers-online FRP games running in Taiwan were developed by non-Taiwanese companies.

The Taiwanese video game industry, without the financial resources of the Korean and Western industries, finds itself technologically backward and stuck with business models that are based on low-cost development. It has been unable for the most part to compete. Individually, video game companies do not have the financial resources to regain the technological lead vis-àvis the Korean industry, and as the industry is widely fragmented, it has been unable to coordinate collective action. A few industry leaders approached III and asked for leadership or help, only to be rebuffed until October 2002. A CEO of one of the biggest companies recounted a tale of a failed attempt to cooperate with III in late 2002:

The worst of the worst is III. A representative of the video game industry approached the president of III, the new one who promised to change III for the better and help the industry, and asked him to help our sector. Mind you, we are the only sector that was really successful in the Taiwanese software industry. After a few talks his answer came back from his secretary: “This is not III core business; therefore we are not interested.” I ask you, “What the hell is their ‘core business’?” They say their goal is to help the software industry, not to make profits, but they do not give a damn about anything but making money. (interview, January 20, 2003)

This failure by III is even more surprising because starting in October 2002 the Taiwanese government declared digital content as a high-priority technological area, specifically targeting online games as one of the more promising digital-content sectors. However, both government officials and industry leaders see the digital-content initiative as a failure so far. According to one of the industry’s representatives in the initiative:

This is the latest government fad. They finally realized IP is important. So we sit there for a year, so many important people and so many people from the government, and talk and discuss for hours every week and nothing happens. It is running for over a year now and still nothing happened. Elections are coming in six months’ time, so nothing will happen until after the elections because the civil servants are waiting to see who will be their master and are afraid to do anything. (interview, October 30, 2003)

This criticism of the progress of the digital-content initiative for online games was echoed by the responsible minister of state, who contrasted efforts so far with government efforts in the hardware sector: “Nothing much is happening with online games. We basically did nothing; we do not even have a budget. We have to admit that a lot needs to be done with software and digital content. However, we [the government] still think that Taiwan’s future is more in hardware” (interview, November 7, 2003).

By the end of 2003, most of the Taiwanese video game companies were already fast retreating from game development. From an industry whose leading companies had operations and sold their own published products throughout the greater China region, the industry was transforming itself to an industry whose core competency is game distribution and resale. This does not mean that two or three of the leading companies will not continue to prosper and grow and develop original games. However, the number of Taiwanese gamedeveloping companies is declining sharply, with no new entrants coming.

In the past few years, a cohort of software companies with new business models have emerged and succeeded. These companies are much more technologically oriented. Their products deal either directly with software technology itself or with new applications of IT technology, like antivirus, opticalcharacter- recognition (OCR), or systems-recovery applications. These companies appeared after the success of the IT hardware industry in Hsinchu Park; indeed, many of them are tightly connected to the industry. There are two types of operations: (a) firms supplying the Taiwanese hardware industry with software technology that enables it to add features to its products, differentiating them from the competition, or supplying critical software that the hardware industry would find difficult or prohibitively expensive to get abroad; and (b) companies producing products that are directly associated with the software industry itself, such as antivirus products or application-development tools.

One prominent example of a company focusing on software technology itself is Trend Micro, which develops corporate antiviral protection products. In 2006 Trend Micro continued to be one of the largest software security companies in the world, with revenues of $618 million USD in the 2005 fiscal year. Trend was established by Taiwanese, grew in Taiwan, is run by Taiwanese, and still conducts a large share of its activities in Taiwan, but its management felt that it needed to leave Taiwan in order to be globally successful. In 1998 Trend moved its management team and headquarters to Japan and reestablished itself as a publicly traded Japanese company; it is also traded on NASDAQ.

The two most successful software companies that followed more closely a business model of alliance with Taiwan’s hardware manufacturers are Ulead and Cyberlink. In 1989 three friends who worked together at III left the institution to try out their own ideas for software development and established Ulead with finance from the Taiwanese scanner manufacturer Microtek. Ulead’s first business was to supply OCR and image-processing software to the fast-growing Taiwanese scanner industry. The scanner industry was facing difficulties in securing critical software from American companies. A few years later, realizing that there was no true-color-imaging processing editor for the PC (Adobe was selling its Photoshop software only for Mac at the time), Ulead launched its own product called Photostyler. In 1992 Adobe bought the company that held the copyrights for the technology Ulead was using. Since then Ulead has come out with its own product for the midrange user.

Hoping to use the same OEM model it used with the scanner industry, Ulead wrote software for video imaging for the video-capture-card industry. The Taiwanese video-capture-card industry never took off, but ironically, this proved to be a boon for Ulead, which started to work with foreign manufacturers. Today, apart from its image-processing products, Ulead also develops video- and DVD-processing and authoring software. Ulead sells either directly to private users or through OEM agreements with hardware manufacturers. In 1999 Ulead became the first software company to go public on Taiwan’s stock exchange. By 2002 Ulead had sales of more than $30 million USD and operations on all five inhabited continents. In April 2005 Ulead, at the same level of revenues, was bought by InterVideo, a Silicon Valley DVDmultimedia software company founded by Taiwanese immigrants. It is striking that one of the three most successful software product companies in Taiwan has had such small total revenues.

The latest globally successful Taiwanese software company is Cyberlink. Cyberlink directly employs business techniques that its CEO, Alice H. Chang, learned as a top executive at Trend Micro during Trend’s rapid growth and IPO period. This is a clear case of diffusion of learning and economic capabilities in the industry. Cyberlink is also the most successful company to be spun off directly from a Taiwanese university lab, attesting to the growing capabilities of Taiwan’s computer science academic research.

Jau Huang, Alice H. Chang’s husband and the cofounder of the multimedia lab at National Taiwan University, established Cyberlink in 1994 together with four of his students.^[29] Encouraged by Chang, then the executive vice president of Trend Micro, the team decided to develop a software product instead of a hardware product. In 1995, using self-financing, Cyberlink was formally founded. The company finished developing its first product, a video (VCD) decoder, in November 1996. In January 1997 Chang stepped in as CEO and embarked on a strategy of OEM sales to Taiwan’s VGA card manufacturers, the world’s largest.^[30] These OEM agreements gave Cyberlink immediate market recognition, and the company went on to develop a complete suite of DVD products that now accounts for about 50 percent of the world market in the PC’s DVD multimedia tool niche. Currently Cyberlink sells multimedia management tools and development tools for electronic training (e-training) and e-learning solutions. In 2000 Cyberlink went public on the Taiwanese stock exchange. In fiscal year 2002 Cyberlink had sales of more than $35 million USD, reaching $75 million by 2005. Most of Cyberlink’s sales originate with its DVD products produced through OEM agreements with PC and notebook manufacturers in the United States, Taiwan, Japan, and Europe.

In short, the Taiwanese software industry today seems to have a dual structure. On the one side are the older private companies dealing with business solutions that need to directly compete with III, which has become the biggest business software solutions provider. On the other side are the younger and more successful companies that are much more technologically oriented, developing in market niches in which III has no dealing, with a few of them beginning to have true global reach. Overall, the Taiwanese software industry is still mostly oriented to the domestic market. Only a few companies possess the necessary capabilities and skills to develop products and services for the global market. Relative to the Taiwanese hardware sector or compared with software in countries such as Israel or Ireland, Taiwan’s software industry is a stark example of a stage 2 failure—the inability of the state to change its roles in tandem with the industry’s growth.

[edit] Government Actions, Institutions, and Market Structure in the Divergent Growth of the IT Hardware and Software Sectors in Taiwan

We can now position Taiwan on our explanatory framework to gain understanding of the institutional systems in which the Taiwanese industry evolved. First and foremost, we notice that the Taiwanese state has opted to deeply intervene in the IT-industry development path by encompassing the R&Dcreating agents within its bureaucratic structure, as well as by trying to intensely influence the industry’s technological development decisions. The Taiwanese state initiated policies with a clear division of labor between private industry and the public sector. The public sector has been conducting the lion’s share of R&D and then diffusing the results, developed to the stage of working prototypes, to the industry. The state also sees itself as the proper locus for making decisions on which R&D projects to pursue and which technologies to develop and obtain. This is almost in direct opposition to the role played by the state in both Ireland and Israel. A recent example is the current system-on-chip (SoC) development program.

The SoC program is the last in a long tradition of technological capability upgrading programs. It is organized in three discrete stages.

1. The government focuses on specific products it thinks that the Taiwanese IT industry needs to be able to manufacture in the midfuture.
2. The universities and ITRI are asked to propose which intellectual property and technologies are necessary to develop these products, and how they are going to develop these technologies and transfer them to the industry.
3. The universities and ITRI implement the development and transfer planned in stage 2.

Hence, in a way similar to the past, the envisioned role of private industry in this innovation capability–development program is limited to one of developing products decided upon by governmental officials using technologies developed by the public sector.

The SoC program also exemplifies the different role for universities envisioned by the Taiwanese state.^[31] Whereas in Israel the universities have for many years been geared toward academic research and have been given wide academic freedom in defining and pursuing their goals, in Taiwan the universities have been viewed mainly as skilled labor–creating mechanisms. This view is especially evident in the new career-path and tenure regulations.^[32] Recently the envisioned role of the universities has somewhat changed, and the government has realized that the university system should be seen as a key agent in the development of new technological and innovational capabilities.^[33] Nonetheless, the overarching technological goals of Taiwan are still defined by the state, and academia is seen as a subsidiary tool that is used to achieve these aims. The universities, therefore, are urged to develop their faculty and departments in accordance with the state’s S&T industrial policies. Part of the SoC program is a detailed development of the technological capability of the university sector, sponsoring eighty-five new faculty positions in predefined subfields of specialization.^[34]

Another example of the state’s direct intervention in the industry’s technological development path is the R&D tax-incentive program of one of the main governmental investment vehicles, the Industrial Development Bureau of the Ministry of Economic Affairs (IDB). This program gives tax incentives and other benefits only to companies that make products that are specified in extensive detail by the IDB itself. The products are gathered in a list based on market research by the IDB every two years. Therefore the list does not include truly innovative technologies whose market is unpredictable or does not yet exist. Moreover, the program specifies that the tax incentives are to be given only to companies that control the whole production chain—that is, from the design to the manufacturing and sales—and only with products that, according to an official, “are not in niche markets that are so innovative that we cannot predict the future so we cannot really help” (interview, May 28, 2001).

Various other government programs finance industrial R&D. In fact, several of our interviewees from the major developmental agencies complained that Taiwan has too many small agencies and programs, each with its own bureaucracy, resulting in a policy hodgepodge and confusion. Of these programs the three that have been most relevant to the IT industry are the SBIR, the Technology Development Program (TDP), and the Leading Product Development Program (LPDP), all of which are sponsored by MoEA Department of Industrial Technology (DOIT). Aside from SBIR, most programs assist mainly in the development of products that are based on some R&D effort, disregarding whether they are new to world markets or not. Most of the grants to IT hardware companies assist them in pursuing second-generation product innovation and do not provide incentives for a riskier approach.

Moreover, although DOIT has been leading the way in fostering more R&D activities in the private sectors in the past three years, it started to channel grants directly to private companies only in 1998. DOIT started to channel a small percentage of the funding formerly given to the public research institutions (ITRI and III) directly to the private industry only after intense lobbying.^[35] Both the SBIR and the TDP are staffed with ITRI personnel who are seconded to the programs. Indeed, according to our interviewees, around one-third of the IT companies that are funded through SBIR also have contract research relationships with ITRI.

These constraints are still more severe in the case of software R&D. For example, thus far software is not even included in the IDB’s tax-incentives scheme. During the interviews many software company executives complained that apart from a brief period in the late 1990s, before the latest government budgetary cuts, they have not been able to get any grants. In addition, the Taiwanese state policy of focusing on specific niches is not appropriate to the current state of the Taiwanese software industry: it denies many promising software companies access to R&D capital, even while hardly promoting the creation of new companies in the targeted niches. Representatives of software companies that did manage to secure state grants complained that the bureaucratic regulations forced them to spend more money on getting the grant than the grant itself provided. Those that did get the grants in the late 1990s,