Revisiting SEMATECH: Profiling public- and private-sector cooperation

The New Competitive Position of the U.S. Semiconductor Industry

The apparent resurgence of competitive strength in the U.S. semiconductor industry has attracted recent attention from scholars and others (Macher et al., 1998; Bridwell and Richard, 1998). This phenomenon is especially interesting in light of the dire predictions made in the late 1980s about the threat of a widespread collapse of U.S. semiconductor firms. The semiconductor industry attracts interest as a bellwether for the future health of the U.S. economy. First, its products are integrated into the products and services of many other industries, adding value to those outputs. As noted in a 1998 report by the Semiconductor Industry Association (SIA) (1998):

In 1996 the chip industry created $41.6 billion of value for the U.S. economy-20 percent more than the motor vehicle parts and accessories industry.. Between 1987 and 1996, the industry's contribution to the U.S. economy grew 15.7 percent per year, more than three times faster than the overall economy grew. The semiconductor industry's value added increased from $11.2 billion in 1987 to $41.6 billion in 1996.. By 2002, the industry's share of the economy will likely double.

Second, due to its knowledge-intensive and rapidly changing nature, the semiconductor industry serves as an example of the emerging competitive dynamics of the so-called "New Economy" (Arthur, 1996). The New Economy is a label popularly used to describe the emergence of a new paradigm of economic growth due to advances in technology and corporate strategy. Thus, the semiconductor industry could be used to construct a general model of industrial competition and growth applicable to other hightechnology sectors.

The renewed dominance of American industry in the global semiconductor market has been documented in several reports. One example is the following passage from a recent report by the nonprofit Council on Competitiveness (1998):

American firms now compete with global strategies that focus on product-driven R&D and partnerships in order to counter the strengths of Japan's vertically integrated industry. American efforts are starting to pay off-the United States regained leadership in global semiconductor market share in 1992 by garnering 43.7 percent of the world market versus Japan's 43.4 percent. This lead has widened.

The exact mechanism driving the resurgence of the U.S. semiconductor industry is too complex to ascribe to a few factors, but recent analyses have identified several trends contributing to the recovery process:

A shift in U.S. firms' technology strategies to emphasize sophisticated, design-intensive devices, such as logic controllers and digital signal processors, instead of competing in the less engineering-intensive dynamic random access memories (DRAM) market.

* The development of a new horizontally specialized industry structure which integrates "fabless" design firms (those with no manufacturing facilities of their own) with the manufacturing capabilities of dedicated semiconductor foundries, which are often located in Asia (Bridwell and Richard, 1998).

* The abundance of venture capital available for investment in new and innovative semiconductor design firms.

* Increased collaboration between U.S. semiconductor firms, and between semiconductor firms and their equipment suppliers.

* Improved cooperation, communication, and research collaboration among semiconductor firms, the federal government, and universities (Rea et al., 1997).

One facet of the U.S. semiconductor industry of interest to academic scholars and practitioners is the possibility that this industry presents a new model for achieving competitiveness. In particular, U.S. semiconductor firms demonstrate an unprecedented level of horizontal and vertical cooperation with other companies, including domestic and foreign competitors, suppliers, and end-users. A frequent criticism levelled at U.S. industry in general in the 1980s was that it emphasized competition over cooperation, when a better balance between the two would have resulted in a higher level of national economic competitiveness. Since that time, there has been broad movement toward increasing collaboration in multiple industries, especially in research and development (R&D). This movement is evident in several areas:

* The relaxation of antitrust laws, particularly through the passage of laws promoting the formation of research joint ventures (Carayannis and Alexander, 1999).

* The formation of research consortia composed of dominant firms in an industry, such as the U.S. Council for Automotive Research in automobile manufacturing and the Grand Alliance in digital television (Vanderdoorpe, 1997).

* The increase in industry-sponsored research at universities and industry-supported university research centers (Cohen et al., 1994).

* The emergence of government-university-industry (GUI) strategic partnerships in R&D to support specific industry sectors, including semiconductor manufacturing (Carayannis and Alexander, 1999).

In view of this trend toward increased research collaboration throughout U.S. industry, and particularly collaboration among government, university, and industry research organizations, it is useful to examine more closely the features of collaboration in the U.S. semiconductor industry and to analyze the effect of that collaboration on industry competitiveness, both in the past and with a view toward the future.

The primary vehicle of collaboration cited by many studies of the U.S. semiconductor industry is SEMATECH (SEmiconductor MAnufacturing TECHnology), the joint venture of semiconductor manufacturing firms formed in 1987 with the approval and support of the U.S. federal government. Other collaborative mechanisms are also important to note, including the Semiconductor Research Corporation (SRC), which sponsors and directs industry-funded academic research to support semiconductor manufacturing; and instances of cooperation between SEMATECH and/or SRC and other organizations. These other organizations include associations of suppliers (i.e., Semiconductor Equipment and Materials International, [SEMI]), government laboratories such as the Sandia National Laboratories in New Mexico, and government-supported university research centers such as the Engineering Research Centers (ERC) program of the National Science Foundation (NSF).

This article begins with an examination of the evolution and management of SEMATECH, which attempts to categorize some of the ways in which the consortium directly contributed to the renewed competitiveness of the U.S. semiconductor industry. It will then examine collaborative efforts between the semiconductor industry and government and university research organizations. The final sections will discuss the new challenges facing the industry as it seeks to maintain its current dominance, rather than catching up to competitors, and the demands that this new situation places on existing and future collaborative ventures.

The Role of SEMATECH in Improving U.S. Competitiveness in Semiconductors

As a pioneering effort in industrial research collaboration, SEMATECH has been analyzed by a wide range of organizations in an attempt to determine its benefits and potential drawbacks. A number of studies (Bridwell and Richard, 1998; Irwin and Klenow, 1996; and Macher et al., 1998) seem to offer a positive overall assessment of SEMATECH's impact on the industry, with some notable exceptions. While it is now widely accepted that there are specific conditions where industrial research collaborations such as SEMATECH have beneficial effects, it is still difficult to identify any ways in which SEMATECH may have supported the renewal of the U.S. semiconductor industry.

Key Developments at SEMATECH through 1996.

SEMATECH was incorporated on August 7, 1987, by 14 hightech companies representing 85% of the national capacity for semiconductor manufacturing. The SEMATECH facility in Austin, Tex., was formally opened in November 1988. The Defense Advanced Research Projects Agency (DARPA) was selected by Congress as the executive agency to manage the appropriated funds earmarked for SEMATECH.

SEMATECH first held a series of workshops to determine those manufacturing areas where SEMATECH should focus. These meetings were an early form of what later came to be termed technology roadmapping. The scope of R&D needs identified in these workshops well exceeded what SEMATECH could hope to accomplish with its $200 million annual budget. SEMATECH's limited budget, amounting to less than 5% of U.S. companies' annual R&D investment in semiconductors, forced the consortium to focus on a narrow slice of semiconductor research issues.

The decision to sell Perkin-Elmer, the leading U.S. manufacturer of lithography equipment, raised further concerns that erosion of the domestic equipment supplier infrastructure would make it impossible for U.S. companies to regain semiconductor market share. Although U.S. companies were competitive in all equipment sectors except for lithography, there was widespread concern that Japan was on the way to establishing a manufacturing equipment cartel. These events led Bob Noyce, SEMATECH's first CEO, to lead a shift in the focus of SEMATECH from horizontal collaboration on new product technologies to strengthening domestic semiconductor manufacturing equipment suppliers by improving their technical capabilities and by being proactive in brokering these suppliers' business strategies.

Although the mix has changed over time, SEMATECH has spent roughly 80% of its R&D funds on activities intended to produce useful results between 12 months and 3 years from the time an activity is started, and roughly 20% has been spent on R&D that will provide useful results in 3 years or more. To do this, SEMATECH has developed advanced processes to challenge and exercise next-generation manufacturing equipment. Emphasis has been on manufacturing equipment and systems technology, not the products, for example, DRAMs, made with that equipment. Therefore, SEMATECH has operated as an industry-wide, vertical consortium focused on strengthening U.S. suppliers for their member companies. The term quasi-vertical refers to the fact that while SEMATECH's emphasis was on strengthening their U.S. suppliers, those suppliers were not members of SEMATECH. Thus, horizontal cooperation was relatively easy to achieve, because SEMATECH's focus was not on their members' products (DRAMs, microprocessors, application-specific integrated circuits, etc.) but, rather, on tie common infrastructure of equipment that SEMATECH's members use to make their semiconductor products.

SEMATECH's work with equipment suppliers includes funding the development of new equipment by suppliers, improving the reliability of manufacturing equipment, qualifying new and improved manufacturing equipment for plant application, developing roadmaps for microelectronics technology and its manufacturing equipment, developing standards for equipment interfaces, and improving communications between semiconductor manufacturers and equipment makers.

Prior to the establishment of SEMATECH, the relationship between the U.S. semiconductor manufacturing equipment industry and U.S. semiconductor manufacturers was much like other vertical relationships in U.S. industry: combative, confrontational, competitive, and often dysfunctional. SEMATECH was established at a unique period in the evolution of these two U.S. industry sectors, during a time when the manufacturing equipment supplier industry was attempting to shift from a dependent relationship with semiconductor manufacturers to an interdependent one. SEMATECH has invested much of their resource base in leading this shift in a way that advantaged their members, not necessarily their equipment suppliers. While vertical relationships in the semiconductor sector are still combative, confrontational, and competitive, because of SEMATECH, vertical relationships are now functional.

This evolution in SEMATECH's role from its founding until 1997 provides some insight into the basis for collaboration among horizontal competitors in research and development. SEMATECH was established to pool resources for the development of "precompetitive" technologies. The member companies originally were unable to agree on what technologies were truly "precompetitive," preventing further progress. Cooperation among members emerged only after "pre-competitive" was redefined to focus on technologies for the semiconductor manufacturing equipment industry, the suppliers to the SEMATECH members. Why, then, did these suppliers agree to cooperate with SEMATECH, a course that only increased the leverage of U.S. semiconductor manufacturers over their equipment vendors?

While SEMATECH was formed in response to the strong challenge to U.S. semiconductor manufacturers by Japanese competitors, this challenge had even more dire implications for suppliers to the U.S. semiconductor industry. Since Japanese firms had their own dedicated networks of suppliers, the decline of U.S. semiconductor firms directly reduced the market for U.S. semiconductor equipment makers. The U.S. semiconductor companies had the option of sourcing equipment from other nations, but their suppliers were essentially shut out of selling to non-U.S. firms. Therefore, areas of research that the equipment vendors might consider competitive under normal economic conditions became "pre-competitive," in the sense that cooperation in technology development became a matter of survival for the whole semiconductor equipment industry.

In effect, a company's definition of "pre-competitive" depends on the company's position in the industry value chain and its market situation. In general, "pre-competitive" technologies are those utilized in the upstream segment of the value chain, such as manufacturing process technologies. Only the threat of a disappearing market pushed the equipment suppliers toward cooperative research. Collaboration with their customers in SEMATECH became an economic necessity for the semiconductor equipment supply industry. Still, cooperation by the suppliers was secured only after SEMATECH established guarantees that technical data and other proprietary information about each supplier's technology and products would be shared only among SEMATECH members, and not with that supplier's competitors. This shows that even economic motives for collaborative research have limited influence over the degree of cooperation between horizontal competitors.

SEMATECH and Its Benefits to the U.S. Semiconductor Industry. Evaluating the role of SEMATECH in the recovery of the U.S. semiconductor industry requires several steps. First, some estimate must be made of the benefit derived by SEMATECH members from the consortium. Second, there must be some indication that "spillover" benefits were produced for non-member firms, either through the members or through other mechanisms. Third, it should also be shown that these benefits would not have been realized without the existence of SEMATECH.

Due to the nature of SEMATECH and its research, it is difficult to derive very exact quantitative estimates of the benefits of the consortium to the semiconductor industry. Most evaluations have used oblique measures to infer the existence or absence of such benefits. Macher et al. (1998) argue that the continued support of SEMATECH by its member firms must mean that the consortium has some value to the industry. Conversely, the drop in the membership in SEMATECH since its formation could be perceived as proof that SEMATECH primarily serves the interests only of its members, and not the semiconductor industry in general. This is also an indirect conclusion, since a number of firms left the consortium due to external factors unrelated to its performance; some firms left the semiconductor manufacturing business entirely, and some merged with other members.

Irwin and Klenow (1996) used firm-level data on SEMATECH members and non-members to estimate the total benefit from the consortium. Their study emphasized the reduction in duplication of research as the primary benefit. The quantitative estimate of this benefit was $300 million over the study period, 1987-1994. As a review of the study by Klette et al. (1999) shows, this estimate omits dynamic benefits from research, so the actual value to member firms was well above $300 million.

SEMATECH members are the dominant competitors in the U.S. semiconductor industry, accounting for the majority of employment and sales in the sector. By definition, any benefits derived by these members affect a large portion of the U.S. semiconductor industry. It is much more difficult to prove the value of SEMATECH to non-members, many of which are small, design-oriented firms. Several aspects of SEMATECH's history indicate that it did help improve the competitiveness of firms beyond its immediate membership:

* SEMATECH moved to fund primarily extramural R&D, thus encouraging innovation across a wider range of firms than if the consortium had funded only internal projects. Spending on R&D outside of SEMATECH rose from 20% in 1988 to 50% in 1990.

* SEMATECH adopted a new mission, starting in 1992, to support explicitly the promotion of an infrastructure of semiconductor equipment suppliers, which served manufacturers beyond the membership of SEMATECH.

* SEMATECH took a lead role in the effort to develop the National Technology Roadmap for Semiconductors, which used input from government, university, and industry to identify strategic areas for R&D investment in semiconductors (Rea et al., 1997). This roadmapping effort helps many semiconductor firms to focus their research on critical issues, and helps the industry as a whole to spot critical technical barriers and to distribute research responsibilities between government, university, and industry performers.

The government's involvement in SEMATECH, motivated by the perceived importance of semiconductors to national economic and military security, necessitated the formation of oversight bodies that combined government and industry leaders. These are new forums for GUI coordination in research, such as the National Advisory Committee on Semiconductors and the Semiconductor Technology Council (formerly the Advisory Council on Federal Participation in SEMATECH). These bodies represent an important evolution in the mechanisms for the coordination and management of national research efforts in semiconductors, which constitute a very broad "spillover" benefit from the operations of SEMATECH.

A significant aspect of SEMATECH in the revival of the U.S. semiconductor industry is its role as a coordinating and focusing organization for the formation and management of GUI collaboration in semiconductor research. SEMATECH's extramural research program and its activities in conjunction with the SIA and the SRC have extended the influence of SEMATECH in setting research directions in various research institutions. These include:

* The Extreme Ultraviolet (EUV) Lithography cooperative R&D agreement (CRADA) between a group of SEMATECH members and the U.S. Department of Energy (DOE),

* SEMATECH Centers of Excellence at major universities throughout the U.S. and SRC participation in the ERC program of the NSF, and

* SEMATECH's support for the Focus Center Research Program (FCRP) of the Microelectronics Advanced Research Corporation (MACRO).

The EUV CRADA

A consortium of semiconductor companies led by Intel has been formed to support EUV lithography research. Called EUV LLC, this consortium's founding members also include AMD and Motorola. Intel has hired Charles Gwyn, Sandia's SEMATECH program manager, to help manage this effort. In 1997, EUV LLC signed a CRADA with three DOE national laboratories (Sandia, Lawrence Berkeley, and Lawrence Livermore). The CRADA funds research into methods for using EUV lithography as the basis for the next generation of semiconductor manufacturing processes.

The project is intended to have the following results (Bonus, 1998):

* Ensure that the technical specifications-i.e., the crucial product standards-for the next generation of lithography tools are set in the U.S. by U.S. equipment producers and their customers (chip-makers).

* Create a significant U.S. player in next-generation lithography tools while adding market competitors to the current duopoly that controls supply of lithography"tools.

* Provide first access for U.S. chip-makers to next-generation equipment.

* Reinforce and amplify the U.S. skill and supplier base in leading-edge semiconductor production tools and technologies.

* Perhaps even create access to the Japanese market by encouraging Japanese chip-making firms to adopt the Americansponsored technology standards.

The EUV CRADA has recently received criticism from members of the U.S. Congress due to the participation of foreign semiconductor equipment manufacturers, such as the Dutch stepper firm ASML Inc. and the Japanese firm Nikon (Bonus, 1998). Also, the Institute for Defense Analyses, a U.S. Defense Department think tank, pointed out in a recent report that the members of EUV LLC have extensive technology alliances with foreign firms and that there is a strong possibility that developments from the CRADA will be transferred to those overseas firms, eroding any military advantage to the U.S. from its next-generation lithography technology. Despite objections from the Congress, the DOE in February 1999 agreed to allow the Dutch semiconductor equipment supplier ASML to participate in the effort through a joint agreement with EUV LLC, subject to certain conditions about technology access and transfer. The DOE emphasized that the benefits of participation by ASML outweighed the risks of overseas "leakage" of U.S. technology, as it provides a liaison between the EUV-DOE collaborative effort and a comparable EUV project in Europe. The European project is partly funded by the European Commission and is led by ASML with the Carl Zeiss Group and Oxford Instruments Plc. The European Commission government is pitching in 10 million European currency units (US$8.7 million) to help fund the project, which will operate within an Esprit program, called the EUV Concept Lithography Development System (EUCLIDES).

The policy debate over the formation and terms of the EUV CRADA highlights the problem that globalization of research presents in the formation and management of GUI consortia. Increasingly, such consortia are forming the infrastructure for developing international systems of innovation, in some cases superseding the priorities of domestic institutions. Due to the international diffusion of technological capabilities, policies that limit consortia to purely domestic players may be shortsighted, limiting access to needed resources to help domestic industry.

The EUV case also illustrates some of the potential risks involved if government participants are cast in the role of selecting "winners" and "losers" in technology. The EUV lithography process is now in competition with a new X-ray lithography method (called Scalpel) from Lucent Technologies' Bell Labs to be the preferred next-generation lithography system approved by International SEMATECH. (International SEMATECH emerged from SEMATECH in 1994 when non-U.S.-based semiconductor firms from Korea, Taiwan, Germany, and other countries were allowed to participate. Japanese firms are still not allowed to participate, as they have formed their own consortium, called SELETE, and they do not observe reciprocity in research collaborations.) So far, there are indications that both technologies will be approved. However, it is possible that EUV will not prove feasible as a solution to next-generation manufacturing challenges in semiconductors, leaving the DOE open to criticism for picking the "wrong" technology.

Semiconductor Industry Research at the NSF ERC & SRC Program

As with the national laboratories, SEMATECH has extensive links to research universities. Since 1988, SEMATECH has funded Centers of Excellence around the U.S., each of which combines the resources of multiple universities and, in some cases, Sandia National Laboratories. Each center focuses on a specific area of semiconductor manufacturing processes or technology, such as X-ray lithography, metrology, and multi-level metal interconnect. The funding for such centers has been modest, approximately $1 million each, with some centers funding research in as many as six universities.

Amore recent form of partnership between SEMATECH and universities is coordinated through the SRC and NSF's ERC program. The ERC program-was developed based on a 1983 study by the National Academy of Engineering, initiated at the request of the NSF director at that time, which recommended the establishment of a new cooperative program with the following two goals (National Academy of Engineering, 1983):

1. Improve engineering research so that U.S. engineers will be better prepared to contribute to engineering practice.

2. Assist U.S. industry in becoming more competitive in world markets.

Establishment of the ERC program was motivated by the perception that significant engineering advances were occurring through the integration of new developments across traditional disciplinary boundaries (similar to what Kodama [1991] calls "technology fusion") and that engineering education in universities no longer prepared students properly for the way in which engineering research was conducted in industry. This required that the centers established by the programs share the following objectives (National Science Foundation, 1997):

* Provide continual interaction of academic researchers, students, and faculty with their peers, namely, engineers and scientists in industry, to ensure that the centers' research programs remain relevant to the needs of the engineering practitioner and that they facilitate and promote the flow of knowledge between the academic and industrial sectors.

* Emphasize the synthesis of engineering knowledge; that is, the research programs should seek to integrate different disciplines in order to bring together the requisite knowledge, methodologies, and tools to solve problems important to engineering practitioners. Contribute to the increased effectiveness of all levels of engineering education.

The ERC program is managed out of NSF's Engineering Education and Centers Division of the Directorate for Engineering. The ERC program issues solicitations for the establishment of ERCs, each with a specific technological focus, such as data storage systems or telecommunications research. Universities submit proposals to host an ERC; these proposals are peer reviewed by technical researchers and executives from academia and industry. Each ERC is funded by the NSF for 11 years. There are 26 ERCs now in this program. Over the course of the 11 years, each center is expected to generate funding from sources outside the NSF, so that the center is self-sufficient by the end of the grant period. To illustrate, in the fiscal year 1994, the 21 centers then in the ERC program received $51.7 million from the ERC program office; $53.7 million from industry in cash, in-kind donations, and associated grants and contracts; and $73.5 million from university, nonprofit, and other U.S. federal government sources.

Each ERC forms several consortia involving university faculty and staff, students, and multiple industrial firms (and on occasion government research facilities as well) to pursue specific research projects under the ERC's focus area. By this structure, projects tend to focus on more fundamental research of broad interest to industry, rather than on development of specific product technologies for individual firms. The academic and industrial researchers exchange knowledge regarding the needs of industry in the research area, relevant developments across engineering disciplines, and the processes of academic and industrial research. The performance metrics used by the ERC program to evaluate individual centers include the numbers of partnerships formed, patents filed and awarded, licenses granted to industry, and undergraduate and graduate degrees granted to students involved in ERC projects.

ERCs are selected to focus resources on complex engineered systems that are several generations ahead of current technology and that will have significant economic impact. Semiconductor technology is one area of focus for the program. One of the earliest centers established under the program-the Center for Advanced Electronics Materials Processing at North Carolina State University, launched in 1988-dealt with semiconductor research. Two other centers were established later: the Center for Low Cost Electronic Packaging at the Georgia Institute of Technology and the Center for Compound Semiconductor Microelectronics at North Carolina State University. Another center is operated under a joint agreement between the ERC program and the SRCthe Center for Environmentally Benign Semiconductor Manufacturing at the University of Arizona.

The joint NSF ERC & SRC center represents a particularly interesting case in GUI collaboration. This center operates under the joint supervision of the NSF's Engineering Directorate and the SRC, with funding from both bodies. The research is conducted jointly between five universities: the University of Arizona; Stanford University; the University of California, Berkeley; Cornell University; and the Massachusetts Institute of Technology (MIT). While the ERC program itself has a long track record of funding and managing the development of successful university-industry research centers, the program's management approach has been modified in the case of this center to accommodate the mode of collaboration favored by the SRC. Drawing on its own experience in coordinating industry-focused academic research, the SRC has assigned "mentors" from its member semiconductor companies to provide important guidance at the project level to the NSF ERC & SRC center. According to one center participant, these mentors typically track progress on research projects on a monthly basis or more frequently, via teleconference if necessary. The SRC believes that this particular management tool provides a better guarantee that the results of the ERC research will suit the needs of industry. Participants in this ERC report that they do have more frequent and substantive interaction with industry partners on a continuing basis and that the mentors also facilitate access to other researchers and managers from the center's corporate partners, enabling more open dialogue on industry needs and priorities.

MARCO: The Focus Center Research Program

MARCO is a not-for-profit research management organization and a wholly owned subsidiary of the SRC. The SRC is a research management consortium that was established in 1982 as the university research arm of the SIA. MARCO has its own management personnel but uses existing support and infrastructure services provided by the SRC (MARCO, 1998a).

In late 1995, James Glaze, then-vice president of technology for the SIA, approached SIA chair Craig Barrett (president and chief operating officer of Intel Corporation) with the idea of pooling industry and government funding to support very advanced university-based research in semiconductor design and manufacturing. According to Dr. Glaze, he and others felt increasing concern that too much of the industry's research was focused on short-term, incremental innovations aimed at very welldefined technical challenges, so that long-term but extremely significant technical issues were not addressed. This trend had been accelerated by recent reductions in U.S. government funding for university research in electronics. A formal proposal was made to the SIA in January 1996, and after extensive discussions among industry, universities, and the government, implementation of the MARCO initiative began in late 1998 (Glaze, 1999).

MARCO is chartered with the establishment and management of a new, university-based focus center research program (FCRP). The FCRP is a new initiative to support pre-competitive, cooperative, long-range, applied microelectronics research at U.S. universities. This initiative receives 50% of its funding from the SIA, 25% from SEMI/SEMATECH (representing U.S.-based semiconductor manufacturing equipment suppliers), and 25% from DARPA of the U.S. Department of Defense. There are over 30 U.S. companies participating in the FCRP. A governing council, consisting of representatives from the three participating organizations, provides oversight of the FCRP. Funding is targeted to be approximately US$ 10 million per year for a fully operational center. If the concept proves to be successful, up to six centers will be funded ultimately. At this point in time, a pilot program will establish two centers for a trial period of 2 years. The technology areas for the two pilot centers will be "design" and "interconnect" (MARCO, 1998b).

The FCRP sponsors developed a joint proposal solicitation to universities. Proposals were submitted to DARPA and reviewed by industry and government evaluation teams. Industry and DARPA program managers produced a combined recommendation of winners, which was accepted by industry members and the DARPA Source Selection Authority. The two lead universities for the new pilot centers are the University of California, Berkeley for the design/test focus center, and the Georgia Institute of Technology for the interconnect focus center. Each lead university works with a team of affiliated universities in their technology focus areas (Exhibit 1).

MARCO's research is intended to complement, not overlap, research carried out at SEMATECH, SRC-funded university research, and other joint research projects of the semiconductor industry (Exhibit 2).

Unlike SEMATECH, an industry consortium with only corporate members, or the SRC, which funds centers in individual universities, the FCRP will support virtual (or distributed) centers that span multiple universities, with participation from broad constituencies of firms as well as contributions from government researchers. This will tap the best expertise across many institutions in order to build the greatest overall capability in a particular technology area. Due to this distinctive structure, the FCRP can tackle technical challenges that are very different from those addressed by the other consortia in the semiconductor industry. The research topics of the FCRP must conform to the following parameters:

* Emphasis on the elimination of barriers via more revolutionary approaches, paradigm shifts, and the creation of multiple options.

* Longer-range (beyond 8 years to commercialization) research.

* Broader, less granular objectives.

* Fewer industrial business practices applied.

* Heavy emphasis on research efforts with faculty, postdocs, visiting scientists, and students (MARCO, 1998a).

The FCRP thus addresses the technology focus outlined in the National Technology Roadmap for Semiconductors but not addressed by previously existing industry resources. As such, the envisioned centers will:

* Concentrate attention and resources on those areas of microelectronics research that must be addressed to maintain the historic productivity growth curve of the industry.

* Strengthen the university research infrastructure and expand its capabilities in silicon-related research.

* Achieve critical mass through relatively large blocks of funding together with the active participation of industrial visiting scientists.

* Provide an optimal balance of creative freedom and targeted objectives.

The transfer of knowledge from the focus centers to the semiconductor industry builds on semiconductor firms' managements' years of experience in the industry in managing consortia such as SEMATECH and the Microelectronics and Computer Technology Corporation. First, rather than depending on only one institution to manage research in a given technology area, the FCRP will create 11 communities of innovation" linking researchers at multiple universities, all of whom can contribute to progress. Second, the research will be long-range, a category of research best carried out by academic researchers, but still linked directly to industry imperatives by setting priorities through the National Technology Roadmap for Semiconductors.

Third, industry and government (DARPA) support can lead to direct interaction between the university researchers and the end-users of the knowledge generated under the FCRP, contributing to a common understanding of user concerns and disseminating new knowledge more widely. Also, DARPA program managers contribute their experience in long-range research management, balancing the more short-term focus of industry members (MARCO, 1998b).

MARCO is not a "partnership" as found in other areas of university-industry interaction. In this initiative, the substantial bulk of research is conducted by academic researchers in the university setting. Industry participants function primarily to provide direction and oversight of the research, evaluate the usefulness of the results for industry, and commercialize any technologies developed in the laboratory. All intellectual property issues are negotiated directly between member companies and each university in the consortium, rather than negotiations being conducted through MARCO. Although this makes transfer technology somewhat more complex, as each firm is likely to license technologies from multiple universities, it facilitates the partnership by removing the need to harmonize the very diverse intellectual property and licensing policies used by each institution. Also, the bulk of technology transfers between the university centers and companies are expected to take place through informal channels, primarily industry researchers who are sent on assignment to work at the focus centers by their companies.

Lessons for Managing GUI Partnerships in Other High-Technology Industries

The activities of SEMATECH have expanded and changed with the continued maturation of existing semiconductor technology and the resulting evolution of the semiconductor industry. Direct U.S. government funding of SEMATECH ended in 1996. This gives SEMATECH more flexibility in its strategy, as it does not answer as strongly to the public policy imperatives imposed by government funding. Recent developments show that policy concerns do still play some role in the consortium's activities.

A few research studies discuss the role of technology evolution in strategic alliances among firms, and they often note that consortia differ in character in mature versus emerging industries. In some cases, consortia involving dominant members of a mature industry can lead to anti-competitive effects, such as the case in which U.S. auto manufacturers used their R&D collaboration to block the development of airbag collision safety systems (White, 1985).

Clearly, the role of consortia must change as an industry evolves, and this is in fact the case with SEMATECH. SEMATECH has made several major changes in its strategy in recent years.

In 1997, SEMATECH for the first time allowed foreign firms to participate in the consortium, through a subsidiary called International SEMATECH. This group includes five Asian and European firms: Philips Semiconductors, STMicroelectronics, Infineon (formerly Siemens Semiconductor Group), Hyundai Electronics, and Taiwan Semiconductor Manufacturing Company. SEMATECH at one point considered allowing Japanese members, but even with the end of government funding, the group felt that such a move would be too politically risky. International SEMATECH was launched to lead the International 300i (or 1300i) program. This is an effort to bring together the technologies necessary to manufacture semiconductors from 300mm wafers. The effort is notable because while past transitions to new wafer sizes were led by single firms (Intel for 150mm and IBM for 200mm), 1300i recognizes that future transitions require costsharing and research among many firms. Also, international standardization in manufacturing processes will help both equipment manufacturers and semiconductor firms to make this transition; therefore, 1300i not only involves International SEMATECH members but also coordinates some work with a similar initiative in Japan, the Semiconductor Leading Edge Technologies consortium (SELETE) (Ham et al., 1998).

In April 1999, SEMATECH announced that it was dissolving International SEMATECH and would instead give the five nonU.S. members full membership in SEMATECH, making the organization a truly global consortium. Foreign members would have full access to all SEMATECH technical programs, including interconnect, front-end processes, assembly and packaging, design systems, and manufacturing methods. Also, SEMATECH announced at the same time a broad agreement with Hitachi Ltd. of Japan to work on semiconductor manufacturing equipment, a move seen by some U.S. equipment makers as an attempt to force U.S. suppliers to compete more directly against foreign suppliers (Lineback and McIlvaine, 1999).

SEMATECH's organizational change reflects broad shifts in the consortium and its industry. A major change is the progressive decline in U.S. membership. National Semiconductor left the consortium at the end of 1998 for financial reasons. Digital Equipment Corporation withdrew at the end of 1999 as a result of its acquisition by Compaq Computers in 1997 and the subsequent sale of its semiconductor manufacturing operations to Intel. Motorola's Semiconductor Group, which has been spun off as a separate company named On Semiconductor, has notified SEMATECH of its intent to withdraw because SEMATECH does not seem to be serving the company's strategic focus. If Motorola finalizes its withdrawal, this would leave SEMATECH with 11 U.S. members and five non-U.S. members. Thus, SEMATECH's U.S. membership reflects the growing concentration in semiconductor manufacturing, dominated by Intel, Advanced Micro Devices, and IBM, as well as the industry's globalization.

At the same time, the U.S. semiconductor industry is still a source of tremendous start-up activity, with the continuous launch of firms specializing in producing new semiconductor designs for application-specific integrated circuits. These companies are often called "IP firms" as they are primarily involved in producing intellectual property, not actual chips, selling their designs to the major semiconductor manufacturers. The cross-disciplinary technical demands of the move toward a "system-on-a-chip," which integrates many devices on a single microprocessor, increases the demand for these small, focused design companies. The growth of U.S. "fabless" semiconductor firms shows how the U.S. remains a major center of semiconductor R&D while semiconductor manufacturing is becoming much more global in nature. In 1990, there were seven fabless semiconductor manufacturers with $1.6 billion in revenues; by 1997, there were 40 such companies with sales of $6.3 billion (Makimoto, 1999).

SEMATECH is forced to evolve by the technological environment of semiconductor manufacturing. Current manufacturing approaches are rapidly reaching the physical limits to "Moore's Law," the famous prediction about the geometric growth in semiconductor chip density. Overcoming those physical barriers will require new manufacturing techniques and systems. Therefore, unlike other consortia in mature industries which tend to invest more in incremental innovations, SEMATECH must devote some funding to radical new science and engineering to extend the technological trajectory of semiconductor manufacturing. This is the major motivation for the creation of new research management structures such as MARCO and the NSF ERC & SRC.

The evolution of SEMATECH and its increased linkages to university and government research performers present some potential lessons to research managers and officials in universities, government laboratories, and policy organizations. These lessons are applicable to the management of future GUI consortia in other industries experiencing rapid technological advances.

Lessons for University Role in GUI Partnerships

Universities have made many important research contributions in the initial, formative stages of most industrial sectors, as well as throughout the evolution of industrial sectors where commercialization is technology driven and products and processes can be understood by a single investigator, e.g., pharmaceuticals and agriculture chemicals. However, in industry sectors where commercialization is driven by making rapid improvements to existing complex products and complex processes, e.g., semiconductors, and in mature industries, e.g., the petroleum industry, universities' records are less distinguished. We believe that this has nothing to do with the excellence of university research, but is a reflection of the incompatibility between the long-view university research culture and the dynamics of competition in sectors driven by the need to make rapid improvements to the next cycle of technology. As we pointed out earlier, when we started this study we fully expected Sandia to have the same problem, and there is much evidence that SEMATECH members and employees had similar concerns.

One pattern emerging from SEMATECH's involvement in GUI partnerships is the move to fund research centers that combine the efforts of multiple universities. For example, the NSF ERC & SRC Center for Environmentally Benign Semiconductor Manufacturing involves not only the two lead universities, Arizona and Stanford, but also MIT and the University of California, Berkeley. The FCRP is even more ambitious, teaming each lead university with nine other universities in the case of the UC Berkeley design and test center and five other universities at the Georgia Tech interconnect center. Lead investigators involved in collaborations with industry are likely to face more situations where they must not only manage research personnel at their own institution, but must also manage "virtual universities" spanning multiple institutions. It also means that universities will need to adopt a more "co-petitive" attitude, accepting the need to team with other universities in some research areas while competing in others.

Lessons for Government Laboratory Role in GUI Partnerships

If central corporate laboratories in semiconductor companies have trouble synchronizing their research with the semiconductor product cycle so that they can deliver results that impact the next design cycle, it is reasonable to expect a government-owned national laboratory to have even more difficulty. GUI collaborations in research suffer from numerous cultural and systemic barriers to effective interaction, such as differing expectations on time horizons, deliverables, and responsiveness (Betz, 1998).

This difficulty can be eased slightly if the national laboratory is working on research that is several product cycles ahead of current needs or if it is working on research topics that would result in what Andy Grove terms "inflection points" (Grove, 1996). Even then the work at the national laboratory must be synchronized with industry research that is pointing toward the particular needs of the product cycle that would accommodate this dramatic change. On the other hand, if a national laboratory is able to overcome these difficulties and become a valued contributor to short-term semiconductor R&D, it can probably support any industry sector. It is for this reason that SEMATECH is such an important benchmark for a national laboratory. It should be pointed out, however, that difficulties identified in supporting this dynamic, intensely competitive industry sector should. not be generalized to other industry sectors, particularly those where the intensity of competition and time compression are less important or commercialization is dominated by new markets or is technology driven.

Most national laboratories, like corporate central research laboratories, have a research-intensive culture that makes it much easier to support industry sectors where commercialization is dominated by new markets or is technology driven. Difficulties arise when a national laboratory attempts to apply management structures and practices developed for technology-driven or newmarket-driven commercialization to industry sectors where commercialization is driven by the need to improve existing products or processes as it is in the semiconductor industry. Of all the national laboratories, Sandia is perhaps the most strongly driven by engineering considerations; therefore, difficulties Sandia might experience in supporting the semiconductor sector are likely to be even more evident in those national laboratories whose emphasis is on science and basic research.

Lessons for Policymakers and R&D Managers

The history of the relationship between government policymakers and industry R&D executives provides some lessons for both on the management of future collaborations.

While the government contribution to SEMATECH was intended to operate under a "sunset" clause, with funding ending when no longer needed, SEMATECH again proved the adage that programs, once funded, are difficult to kill. Republican members of the U.S. Congress tried very hard to eliminate federal funding to SEMATECH in 1994, but were ultimately unsuccessful due to the institutional alliances SEMATECH had formed over time. This ultimately proved beneficial, as SEMATECH was able to make the conscious decision to phase out direct government funding of its efforts in an orderly manner. The U.S. government continues to support SEMATECH through cost-sharing on various joint research projects, but the experiences of Sandia National Laboratories show that the government receives substantial benefits from such funding.

A more difficult question is how national government should address the internationalization of industry consortia, and particularly GUI collaborations. The formation of the I300i initiative and the international implications of the EUV CRADA are indicative of the kinds of policy challenges involved in government support for future research collaborations. Industrial research is not likely to become less global in nature as time goes on, and the treatment of international linkages in publicly funded research must recognize this reality.

GUI collaborations are best analyzed both at the microscopic level, examining each collaboration individually, and at the macroscopic level, looking at the diversity and scope of interactions between a given industry and the university and government research systems. Raymond Kammer, director of the U.S. National Institute of Standards and Technology, has called this macroscopic view "topsight" in relation to governmentindustry interaction (Kammer, 1998). This view looks at the totality of interactions between research organizations from different sectors and sees how multiple collaborations complement each other in support of industrial competitiveness. The process of technology roadmapping aids in the coordination of multiple collaborations, as different forms of GUI collaboration and different management structures are needed, depending on the type of technology development needed to advance an industry.

Due to the different research strengths of various types of research performers in government, universities, and industry, corporate R&D executives must consider the full range of potential cross-institutional structures for managing collaborative research and determine which particular structures are most suitable for each industrial technological objective and each industry's environment. Exhibit 3 presents one way of portraying the different structures.

Collaboration with universities is very useful in "sciencebased" innovation, but national laboratories may be more appropriate in engineering-intensive efforts. One clear need for joint effort between government, universities, and industry is further study into which configurations are most appropriate for various levels of industrial competitiveness. Such an analysis would help to determine which structures should be emphasized given an industry's evolutionary state, technological trajectory, and competitive environment.