Luncheon Address: The Industry Perspective on Semiconductors
George M. Scalise
Semiconductor Industry Association
Mr. Scalise prefaced his talk by addressing the unfortunate events of September 11 and their impact on the semiconductor industry. Then he offered a quote from Alan Greenspan, chairman of the Federal Reserve Board, made a year and a half earlier: “An economy that 20 years ago seemed to have seen its better days has enjoyed a remarkable run of economic growth that appears to have its roots in the ongoing advances in technology.” He agreed, saying that that observation was still valid and would continue to be valid for a long period.
THE NEAR-TERM OUTLOOK
He began with the near-term outlook, showing a chart on which worldwide semiconductor revenues were down about 37 percent as of July 2001. He called that decline “dramatic” but said it was slightly better than it had been on a month-to-month basis in the prior period. He said that the industry seemed to be dealing with a cyclical issue largely inventory-driven, rather than a structural change in industry. One of the reasons for the decline, he suggested, is related to the investment in capital equipment over the last several years. In the year 2000 worldwide investment in capital equipment nearly doubled, from $25 billion to $48 billion, over the preceding year—“a very large increase.” The forecast for the year 2001 was $31 billion, still roughly 20 percent higher than the investment of 1999. He said that this suggested considerable “excess investment,” most of which had gone to the foundries.
He then compared the integrated device manufacturers12 with the foundries.13 The IDMs traditionally invest about 23 percent of sales to maintain a 17 percent compound annual growth rate (CAGR). Capital investment for the industry as a whole was then well above that range, which he attributed to heavy spending on new plants by the foundry groups—a trend that he said would soon come “back into balance.”
Another “piece of good news” that might bring the industry back into balance in the relatively near term is the book-to-bill ratio for equipment, as opposed to that for semiconductors. Although the ratio had been unfavorable (below unity) since February 2001, it had more recently been moving steadily closer to unity.
Capacity utilization rates, at an average of about 72 percent, were unfavorable and not showing improvement. However, he said, if the companies were broken down into IDMs and foundries, the rates for IDMs were 82 to 84 percent, whereas the foundries were between 30 and 40 percent and had not yet stabilized.
Prices and Inventory
A point that had been discussed late in the morning session concerned functionality and price, measured by the chip price/performance index. This index comprehends not just microprocessors and memory, but a broad spectrum of products. The index is moving at a rate slightly below what Moore’s Law would “require,” which seemed to reflect the excess manufacturing capacity and the resulting price pressures. However, Mr. Scalise said that prices were beginning to return to the 30-percent-per-year rate of decline of recent years. He turned then to excess inventory, which at the beginning of 2001 was at about $15 billion worth of product. That amount had been reduced by nearly half during the year, while demand had grown—“a combination that bodes well for the outlook from here.”
A Turn in the Cycle?
One piece of justification for that optimism was that consumption had begun to exceed shipments for the first time since about April 2001, which was beginning to bring down the inventory in the channel. In the view of the SIA, that meant that the September quarter would be the last down quarter of the current semiconductor cycle, and that the December quarter would be the first growth quarter of the next cycle. As a consequence of September 11, he said, the Septem-
ber quarter would be a little lower than anticipated, and the growth in the December quarter would be a little less. Fourth-quarter growth had originally been projected at 5 to 7 percent over the third quarter, and that projection had been lowered to 1 to 5 percent. “But what is important,” he said, “is that we are still on that same track.”
He turned then to several “megatrends” for the industry. The first was that about 70 percent of all semiconductors were used for communication and computation. These products would be needed in the immediate future to deal with the “terrorist world that we now face”; the government would need hand-held devices for communication and computation, numerous kinds of GPS devices, and faster core memory for transportable PCs. Wireless communication, in particular, was projected to become a much larger market. He concluded that the near-term outlook was not quite as robust as it was before September 11 but still had the “same dynamic.” He referred to a large number of new semiconductor-based products due on the market by the Christmas shopping season, including Black-berry-based devices and miniaturized digital audio products, some of which would replace portable CD players.
For the longer term, he said that information technology (IT) would be the number-one driver of the U.S. economy. While IT represented about 8 percent of the economy, he said, it had provided 30 to 35 percent of economic growth over the previous 5 years. It had also lowered inflation by about 0.5 percent during that period and helped to nearly double productivity.
He said that a shift was taking place in the pattern of semiconductor consumption. In 1995, 56 percent of semiconductors went into computers and 14 percent into communication devices; in 2001 those figures had changed to 46 and 24 percent, respectively, with communications growing rapidly. Consumer products had held steady at around 15 percent. In response to a question, he said that the auto industry consumed about 5 percent of semiconductors.
SEMICONDUCTORS AND STRUCTURAL CHANGES
Impact on the Economy of Computers and the Internet
He described a change in the economy in the mid-1990s, when a host of technologies—including the Internet and e-commerce—converged. Since then the Internet had become more functional, PCs faster, operating systems more reliable, memory and microprocessors cheaper, communications better, and hand-
held devices more numerous. All have converged, he said, to bring new elements that will impact the economy for the long term, in terms both of growth and of making major contributions to the deflationary forces in the economy.
To illustrate this change, he showed the changes in the average retail desktop PC from August 1995 to 2001. It had become 18 times as powerful, its clock rate had risen 12 to 15 times, disk storage had increased 50 to 60 times, and memory had increased more than a dozen times. Computers now had a new dimension of connectivity, beginning with built-in modems, and the price was about half of what it had been. That is the real contribution of this industry, he said, and under Moore’s Law it may be able to continue this contribution for 10 to 15 years. If the industry can move beyond scaling, he said, the impact on the economy will continue to be as strong as it has been for last several years.
The Value of Free Trade
Continued advances will be driven by free trade and globalization, he said. The trade agreements over last three decades, for which the SIA lobbied hard, had “made trade work.” In particular, the U.S.-Japan Semiconductor Trade Agreement of 1986 successfully addressed two major issues: dumping and market access. Foreign suppliers’ share of the Japanese market had grown from 8 percent in the mid-1980s to between 30 and 35 percent today. U.S. suppliers’ share of the world market had recovered from a low of about 34 percent in the mid-1980s to about 52 percent.
A more recent international agreement saved U.S. companies about $1.5 billion in tariffs.14 “Convincing the rest of the world to adopt that treaty was not easy,” he said, “but eventually even the Europeans agreed that tariffs are just a cost that adds nothing to the ability to compete in the markets.” Another result of trade agreements was that innovators were allowed to reap the rewards of their work.
A major international event, he said, is the entry of China into the WTO. As a market for semiconductors, the Chinese have moved into the third or fourth position, and in the next decade are expected to be number two. They are expected to become second in the world in manufacturing as well, at around the same time.
Legal Protection for Bits
Urgently needed in both international trade and e-commerce, continued Mr. Scalise, is an agreement on legal protections covering all digital products, including software, that are as effective as the legal protections covering physical goods. The key to this transformation, he said is to achieve “technological neutrality” so that “bits and atoms” are treated equally under tax and tariff regula-
tions. “Our view is that a good is a good, whether you buy it in a package or download it from the Internet. That’s going to be an important principle for e-commerce.”
Major R&D Needs
Turning to the R&D needs of the industry, he saw a “big problem.” He said that federal spending on basic research as a percentage of the overall federal budget had declined by about 60 percent since 1992. While overall R&D spending had remained nearly constant at about 2.5 percent of GDP, the mix had changed dramatically. The federal share had dropped from a high in the 1960s of about 2 percent to about 0.8 percent, while the industrial share had increased from about 0.5 percent to nearly 2 percent.
He expressed concern that even though the total is roughly the same, it represents “a much less stable investment environment.” That is, during economic downturns, companies tend to cut their investments in R&D, whereas a government can maintain a more even funding rate.
He noted that among all the disciplines mathematics and science had suffered large cuts, while the medical area was receiving large increases.15 He noted that medical advances depend not only on health research, but also on diagnostics, metrics, instrumentation, and other fields that are built on a foundation of basic physical science and technology.
He noted challenges in the area of workforce trends as well. He cited the example of TIMSS scores, in which U.S. students are performing below world levels, and said that much more needs to be done for K-12 students in mathematics and science.16 The number of bachelor’s degrees awarded in electrical engineering has fallen about 40 percent over the last several years. While the decline
seems to have flattened out, the numbers are still disappointing. “This industry is growing fast, and it is 30 percent of the economy, so we can’t afford to have fewer engineers available.” He showed a chart indicating that the number of engineering graduates in other parts of the world is higher than in the United States. In Europe, for example, the number is 140,000 vs. 60,000 in this country. “That,” he said, “is going to be a competitive problem.”
A Call for Federal Research Funding
To address this challenge, he suggested, federal funding needs to support more IT research. In 1998, 30 percent of industrial R&D went toward work in IT, but only 6 percent of federal R&D was done in IT. “If it’s having a 30 percent impact on the economy, we have to invest more at the federal level. That’s the number-one message: Invest in the basic research of math and science that will be the driver of the economy one and two decades from now.”
Growth and Prices
He then turned to the industry as a whole, which historically has grown at a 17 percent compound annual growth rate. Although that figure had dipped during the downturn, he predicted that the industry would keep to that pace for the next 10 to 15 years—“as long as we make some of the key investments we’ll have to make.”
He again referred to Moore’s Law, saying that the price per bit is continuing to remain on track. A bit of memory cost about 5 percent of what it did in 1995 while providing more functionality. The combination of 17 percent growth and 30 percent decline in prices, he added, translates to about a 55 percent compound annual growth rate in the number of transistors produced. In the year 2000, about 60 million transistors were produced for every person in the world, a figure projected to rise to 1 billion by 2008. He suggested that this would translate into real benefits for countries with poor infrastructure, where enhanced wireless communications would allow people much greater capabilities in not only communication but also computation, health care, and other fields.
On the horizon, he said, the roadmap indicates some substantial technological problems—most likely arising between 2005 and 2008—if the industry is to maintain the pace of Moore’s Law. Most of those roadblocks will begin to emerge as the feature sizes of chips shrink to the range of 100 to 70 nanometers; we are now in the 200-nm range. He said that current plans, which include more investment in R&D, should be sufficient to address these challenges.
RESPONSES BY INDUSTRY
Precompetitive Research by Industry
The first plan is the precompetitive work being funded by the Semiconductor Research Corporation. This program spends $40 million per year on directed research carried out in universities by over 700 graduate students worldwide. In addition, International SEMATECH, in close partnership with its member firms, is supporting research on manufacturing technologies at a level of $140 million annually. Third, the Focus Center Research Program, created in 1998, sponsors a multi-university effort to address major basic research challenges identified by the ITRS.
There are now four Focus programs: a design and test program, led by the University of California at Berkeley; an interconnect team, led by the Georgia Institute of Technology; a circuit systems and software program, led by Carnegie Mellon; and a materials and devices program, the most recently formed, led by MIT. Each program has seven to eight partners, and funding for the four programs totals $22 million a year, which will grow to $60 million a year over the next several years. These programs are designed to solve the problems that are already visible in the short term. “We think we can handle the next several years,” said Mr. Scalise. “The issue is what we can do beyond that in the way of basic research.”
Policy Suggestions from SIA
He summarized the SIA’s suggestions under three categories, as follows:
To ensure adequate Federal funding for basic university research, the SIA proposes:
A Federal R&D strategy to guide funding, including a commitment to organizations that fund basic research, such as the National Science Foundation.
A commitment to strengthening research infrastructure to ensure that university labs are adequately equipped and staffed.
To retain the competitive advantage we gain by employing the best and brightest engineers, the SIA proposes:
Fundamental reform of the H-1B visa process to exempt those who earn master’s and Ph.D. degrees in the hard sciences at U.S. universities.
To ensure that the evolution of the Internet is not hindered, the SIA proposes:
Uniform national and international standards for all public policies, including issues such as taxation, tariffs, access, and privacy.
In addition, the SIA has proposed workforce and education initiatives, a partner-
ship with SECME, and a minority pre-college success program that provides professional development for teachers.17
Other SIA Initiatives
The SIA also sponsors a program for technicians with the Maricopa Advanced Technology Education Center (MATEC) to support community college programs for developing fab employees, including undergraduate program assistantships and a program for foundry employees who want to learn design. The SRC has provided support through contract research for master’s and Ph.D. students since its founding in 1982.
He described the industry’s environmental health and safety emphasis, which has earned it a number-two ranking out of 208 manufacturing industries, according to the Bureau of Labor Statistics. The number-one industry, which assembles communications equipment, does not work with the chemicals and gases that are part of the semiconductor industry. The objective of the semiconductor industry is to move to the number-one ranking, and, more broadly, for all companies once a year to share best practices and to achieve the same worker health standards worldwide.
A POSITIVE OUTLOOK
He closed by stating that the outlook for the industry remains positive. “In our view, it continues to be as strong today as it was 10 to 15 years ago.”
In response to a question about funding, he added that the SRC has minimal government involvement—by design. The industry, he said, felt that it needs to provide the funding for the SRC because the research is mainly precompetitive. It has also been internationalized because investments benefit companies around the world. The SRC would like more industrial support, not more federal support; federal funds, it feels, should be devoted to basic research.18 He also said that funding for the Focus Centers is shared among the SIA (50 percent), the equipment companies (25 percent), and DARPA (25 percent) for the support largely of crosscutting technologies, such as metrology. One option is to include other industries in this support, since the research is often broadly based.