In This Issue
Celebrating Manufacturing Technology
March 1, 2005 Volume 35 Issue 1

Semiconductor Manufacturing: Boom Busts, and Globalization

Tuesday, March 1, 2005

Author: Alfonso Velosa

To prosper in the global marketplace, semiconductor companies are developing new business models.

The semiconductor industry presents something of a conundrum to the outside observer. Products based on amazing technologies are developed in huge, sophisticated, very expensive facilities. The majority of these products are then sold for a few cents or a few dollars each in mass markets. This market approach is driven by business decisions that (1) maximize profits and minimize risks in a high-technology market that is extremely capital intensive and subject to severe industry cycles and (2) respond to increasing globalization that started more than 20 years ago. In this paper, I analyze these trends and identify the major business models semiconductor companies have developed to profit in this environment.

Semiconductor Product Markets
The electronics and semiconductor industries, key building blocks of the global economy, combine macro- and microeconomics in many ways. To understand the importance of these industries, consider the size of the global economy in relation to the electronics industry. According to current estimates, the global gross domestic product (GDP) was approximately US$33 trillion in 2003, and the U.S. GDP was approximately US$10.4 trillion. Electronic equipment sales worldwide were estimated at US$1.04 trillion in 2003, and if one includes software, services, and telecommunications, global spending for information technology (IT) was US$2.26 trillion, or about 7.5 percent of total global economic activity.

Semiconductors are the basis for the electronics and IT industries, the crucial building blocks for key elements. The semiconductor industry reached US$178 billion in 2003 and is estimated to have reached US$223 billion in 2004. Not only is the semiconductor industry large, it also represents 15.5 to 21.5 percent of the value of the electronics industry, depending on where it is on the industry cycle in any given year.

Since its beginning as a laboratory experiment, the semiconductor industry has grown to more than $200 billion, employing slightly less than half a million people worldwide. This exciting increase has taken place in many technology areas, involves many companies, and has undergone numerous business cycles, which have regularly shaken up the industry and caused it to change its competitive structure and the way management addresses the market and invests in new facilities and technology. In the past 20 years, the cumulative effect has been to compress the business cycle, thus necessitating short-term planning at the corporate level.

The pressures of the semiconductor business cycle have been compounded by a decrease in the industry’s core growth rate, from approximately 15 to 17 percent through the mid-1990s to 10 to 12 percent in the past decade. The fundamental change in expectations for the industry has complicated risk management, as executives assess investments for new facilities. When the cost of a fabrication facility exceeded the billion dollar mark, executives were forced to develop new business models to minimize their risk.

The cyclical nature and slow growth patterns of the industry also reflect the evolution of core electronics markets. As electronic equipment proliferated and reached the trillion-dollar sales level, it reached the limits of market penetration and brought the rate of growth closer to that of overall global economic growth. The core electronics markets also illustrate the “bullwhip” effect on the supply chain, whereby small changes in orders, say, for computers by end users propagate through the system affecting retailers, distributors, manufacturers, and so on. Eventually, these changes have large, volatile effects on the suppliers of semiconductors, wires, and other building blocks. Thus, the business cycles in the computer, telecommunications, and consumer electronics industries have strong effects on the business cycle of the semiconductor industry. In the debacle of 2001, for example, the semiconductor industry shrank by more than 30 percent.

These end-market shifts reflect the manufacturing shift towards Asia, as firms search for integrated electronic supply chains and lower manufacturing costs. In 2000, the consumption of semiconductors by region was roughly equal among the Americas, Asia, Europe, and Japan. Beginning in 2001, however, consumption shifted towards Asia, and expectations are that Asia will eventually constitute more than 50 percent of the market.

Semiconductor Device and Process Technology
As electronics markets adapt to shifting business cycles and changing geographic manufacturing patterns, semiconductors remain the key technological element in electronic systems. Researchers have developed a technology spectrum for maintaining a full semiconductor pipeline and business processes and intellectual property rights regimes for the next 10 years.

Let’s look at a few of the initiatives for the near term, midterm, and long term. In the near term (less than two years), products that will reach their critical market volumes include silicon-germanium chips, radio-frequency complementary metal-oxide semiconductors (CMOS) chips, and platform application-specific integrated circuits (ASICs). Multiple competing firms are bringing these products and technologies to market, and customers are starting to integrate them into electronic systems. In the midterm (two to five years), technologies such as embedded field programmable gate array (FPGA) cores and organic light-emitting devices will address significant customer issues and will attract funding for product development. In the long term (five to 10 years or more), technologies such as micro-fuel cells, gallium-nitride (GaN) devices, and polymer memory have shown significant promise. However, these technologies must overcome many barriers, such as lack of a supply chain infrastructure and the absence of demonstrated large-scale products, and they may require further development before they can be integrated into high-volume manufacturing facilities.


In the past decade, the
industry’s core growth rate
has dropped from 16 percent
to about 11 percent.

This 10-year time scale is critical to the industry because semiconductor companies are increasingly adopting smaller scale manufacturing process technologies. The largest and most capable semiconductor manufacturers are developing products based on small line-width technologies with the objective of reaching the market ahead of their competitors. New forecasts for semiconductor fabrication facilities by these leading-edge companies show them to be two to three years ahead of the rest of the market, and their lead appears to be growing. Thus, Moore’s law continues to reflect and influence decision making in the semiconductor industry.

The key metric for the industry is feature size or line width (i.e., the size of the wires embedded in the silicon to channel electrons). Leading-edge companies, such as Intel and TSMC, implemented 90-nanometer production facilities in 2004; 65-nanometer processes are expected to be in volume production by 2006. The wholesale adoption of new technologies continues to shift the industry’s center of gravity towards smaller and smaller process technologies, forcing leading older facilities to close as they become less and less cost effective. However, companies that move away from the cutting edge of high-volume manufacturing are finding business models that continue to leverage older (500-nanometer and larger) processes for all electronic markets.

New Business Models
As companies invest in new fabrication facilities to produce new technologies, the risks inherent in these business investments continue to increase. In 1983, it cost $200 million to manufacture 1,200-nanometer chips in a leading-edge semiconductor facility. In 1997, it cost approximately $1.3 billion for a 350-nanometer facility, and in 2003, it cost as much as $3 billion for a 130-nanometer facility. Because of the large fixed costs of fabrication facilities and the significant volumes and capital resources necessary to justify them, the number of firms with leading-edge technology has been restricted to the largest semiconductor firms in the world. By 2003, even firms in the top 10 had to consider carefully where and how to invest in facilities (Table 1).

Tier 2 and 3 firms must have clear strategies or well defined market niches to compete in this arena. The decision to build a new facility or set up a laboratory must be based on business factors, such as which city or region in the United States or Germany or China offers the most money and the best incentives, what the markets are demanding, and the cost of capital to build a $1.5 to $3 billion facility.

The semiconductor foundry market, a model developed in the past decade, is made up of specialized contract manufacturers (foundries) that perform wafer-fabrication services for semiconductor companies that do not have fabrication facilities, integrated device manufacturers, and original equipment manufacturers. In the majority of cases, the customer creates a design for an integrated circuit and transmits it to the foundry, where it is fabricated on silicon wafers and delivered to the customer or another contract manufacturer for packaging, assembly, and testing. By 2003, the center of gravity of the foundry market was in Asia with TSMC with $5.9 billion and UMC with $2.5 billion in revenue, and IBM positioned as a strong technology developer at $0.8 billion in revenue.

The emergence of the foundry business model facilitated the rapid increase in “fabless” start-up companies (i.e., companies that do not have their own fabrication facilities), which no longer had to raise enough capital to build a facility. By definition, the fabless sector is highly dynamic, with many start-ups, many failures, and a few companies fortunate enough to hit on a successful product or emerging market and “ride the wave” as long as they can. Predicting which fabless companies will be winners is all but impossible. Nevertheless, the fabless model is successful, and the growth of fabless companies as a whole is likely to continue and to outpace the general semiconductor market. By 2003, six fabless semiconductor firms had exceeded $1 billion in sales (Qualcomm, NVIDIA, Broadcom, Xilinx, ATI Tech, and MediaTek Inc.)

The intellectual property industry complements the integrated, foundry, and fabless industries, providing solutions to issues related to specific products. Intellectual property has been developed for specific process technologies for chip modules, such as the compute engine, logic, input/output, memory, and so on. By 2003, the intellectual property market had reached critical mass, with several firms close to or exceeding $100 million in sales (ARM-$175 million, Rambus-$118 million, and Synopsys-$82 million).

For several decades, the semiconductor industry has been making significant adjustments to adapt to globalization, providing an instructive example for other industries contemplating full-scale globalization. Both semiconductor production and consumption have become global activities, with centers of excellence spread across the world. The coveted advantages of having a modern semiconductor facility in one’s community has led local governments around the world to provide semiconductor firms with increasing financial and infrastructure incentives to build new facilities in their locales. Semiconductor firms seeking to minimize their risks for these investments in large facilities are negotiating extensively with flexible, responsive local governments.

At the same time, semiconductor firms continue to develop a technology pipeline that complements the current infrastructure, as well as newer approaches, such as the fabless and intellectual-property business models. The flexibility shown by the industry in the pursuit of end markets, collaborations with partners and governments, and the leveraging of technology are grounds for optimism about the future of the semiconductor industry.

TABLE 1 Top 10 Semiconductor Firms (in US$ millions)

2003 2000 1997
Intel $27,103 Intel $30,298 Intel $21,746
SAMSUNG Electronics $10,502 Toshiba $10,866 NEC $10,222
Renesas Technology $7,936 NEC $10,643 Motorola (Freescale) $8,067
Texas Instruments $7,410 Samsung $10,585 Texas Instruments $7,352
Toshiba $7,356 Texas Instruments $9,202 Toshiba $7,253
STMicroelectronics $7,180 STMicroelectronics $7,890 Hitachi $6,298
Infineon Technologies $6,864 Motorola (Freescale) $7678 Samsung $5,856
NEC Electronics $6,312 Hitachi $7,286 Fujitsu $4,622
Motorola (Freescale) $4,628 Infineon Technologies $6,732 Philips Semiconductor $4,440
Philips Semiconductors $4,513 Micron Technology $6,314 STMicroelectronics $4,019
Source: Company reports, 2003.

About the Author:Alfonso Velosa is associate director of marketing and business strategy at Gartner Consulting.