Advanced Micro Devices, Inc.

views updated

Advanced Micro Devices, Inc.

901 Thompson Place
P.O. Box 3453
U.S.A.
(408) 732-2400
Fax: (408) 749-3375

Public Company
Incorporated: 1969
Employees: 11,250
Sales: $1.23 billion
Stock Exchanges: Boston Midwest New York Pacific

Advanced Micro Devices is the fifth-largest maker of integrated circuits and microprocessors in the United States. It also produces memories, programmable logic devices, and circuits for networking applications, office automation, and telecommunications.

AMD was founded by Jerry Sanders, who had quickly risen through the ranks of Fairchild Semiconductor to become the firm’s marketing director at the age of 31. In 1969 Sanders and eight other employees left Fairchild to form AMD with $50,000 of their own money and $1.5 million in venture capital, far less funding than other semiconductor start-ups like Intel.

With so little capital, AMD did not have the resources for research and development, so it had to begin making money quickly. As a result, the firm initially acted as a secondary source of chips, manufacturing chips designed by other firms—Fairchild and National Semiconductor initially, and later Intel—under license. At that time second-source companies rarely came out with their own products, instead serving primarily as an emergency source of chips during shortages and acting as a guarantee that primary-source companies would not overcharge the manufacturers that depended on their semiconductors. Given this niche, Sanders devised several strategies to pull his firm out of the pack.

AMD was one of the first firms in Silicon Valley to stress quality above all else, designing its chips to meet United States military specifications for semiconductors. At a time when the young computer industry was suffering from unreliable chips, this gave AMD an advantage, and the firm catered to customers in the computer, telecommunications, and instrument industries who were growing quickly and who valued high reliability enough to pay more for it. AMD avoided producing chips for inexpensive consumer items like calculators and watches, feeling they were only short-term markets, even though catering to these markets might have resulted in a needed infusion of cash.

The firm used price incentives before its competitors, and Sanders also relied heavily on salesmanship to keep the company afloat, pushing his sales force to simply outsell the competition. To do this, he kept the company decentralized, breaking it into several product profit centers. Each had a managing director who was directly responsible for marketing and pricing, as well as product and technology development, and worked directly with customers. As a result, engineers and designers were more aware of the business implications of their work than were their counterparts at most Silicon Valley firms.

A flamboyant leader who flaunted his love of materialism, Sanders used his personality to push his small company into the public eye, giving it a larger presence than its size merited. While attempting to motivate employees through the desire to be as rich as he was becoming, Sanders stressed respect for those low on the company’s totem pole. He threw extravagant Christmas parties for everyone in the company and one year held a raffle, awarding $12,000 a year for 20 years to the winning employee—and showed up with a camera crew to record the prize delivery. These practices contrasted markedly with those of AMD’s more conservative competitors like Intel and National Semiconductor and quickly gave the firm an aggressive reputation.

AMD went public in 1972, and by 1977 was one of the fastest-growing semiconductor manufacturers. The company had reached $100 million in annual sales, but remained in the secondary manufacturing market. Tired of paying royalties and seeking to expand its horizons, AMD signed an agreement with Siemens, a huge West German electrical firm. Siemens bought 17 percent of AMD’s common stock, and the two firms set up a joint venture to build and sell micro-computer systems. The deal raised $22.5 million, with which AMD intended to start designing its own chips.

The money proved inadequate to the task, however, and AMD pumped ten percent of sales into research and development. The firm’s debt expanded to $29 million, despite sales growth to $225.6 million by 1980. One of the hottest markets in the computer industry was microprocessors, which duplicated the functions of early computers on a single chip. However, Intel held a commanding lead in this area, and AMD once again acted as a second source for microprocessors made by Intel and Zilog Corp., an Exxon subsidiary. AMD did develop its own devices to link microprocessors to larger computers.

To design new products, AMD used a different approach from that of most other semiconductor companies. Rather than develop innovations in isolation from the market, AMD employees were instructed to look for new or unmet needs by surveying potential customers and documenting their comments. When important needs were located, engineers were assigned to design chips to address them. This approach led to chips designed to send computer data over telephone lines and link office telephones and switchboards.

During the late 1970s and early 1980s, Sanders pushed hard to try and become a major supplier of proprietary devices to the telecommunications industry, drawing knowledge from Siemens, which was already involved in the European telecommunications industry. The United States telecommunications market proved hard to infiltrate, however, since the telephone companies made much of their electronics themselves. Nevertheless, AMD was clearly successful in the area of bipolar chips, earning half its revenue from them by 1982. These profitable, proprietary chips were faster than most other types of integrated circuits and in high demand throughout the computer industry.

The personal computer (PC) market was beginning its explosive growth, and competition for engineers was intense. Partly for this reason, AMD introduced a no-layoffs policy in 1980. The move increased company loyalty, and kept turnover at about 20 percent of the industry average during the early and mid-1980s.

As the telecommunications and computer industries expanded, the semiconductor industry grew rapidly, reaching $7.8 billion in 1981, of which AMD’s share was about $400 million. The industry became so competitive and development costs so high that AMD and rival Intel signed a ten-year agreement in 1981 calling for technology exchanges and cross-licensing in an attempt to hold down their costs. As a result of the pact, AMD moved heavily into the manufacture of the Intel 8086 microprocessor family, used in the rapidly growing PC market. AMD had become the seventh-largest United States chip manufacturer, and was growing faster than Intel, expanding by 17 percent in 1982, as opposed to Intel’s 14 percent. Intel was still three times the size of AMD, however. Despite the deal with Intel and its success with bipolar chips, AMD continued to lag in metal-oxide semiconductors (MOS), the dominant chip-making technology. The company poured research and development money into MOS technology, hoping to catch up to Intel and other firms.

From 1972 to 1984, AMD grew faster than any other Silicon Valley semiconductor manufacturer, in part because of Sanders’s emphasis on quality and focus on market needs. The firm made 550 different chips and had a diversified customer base with no buyer accounting for more than ten percent of sales. Customers included AT&T, IBM, Digital Equipment Corp., Burroughs, Nippon Electric Corp., and Olivetti. In many cases AMD was still a secondary source, licensing its designs from other companies, but about 40 percent of sales came from chips designed in house.

AMD was the second-largest manufacturer of bipolar chips after Texas Instruments. Its high-performance bipolar chips were being used in large computers, radar, and highspeed graphics. In contrast, Intel was not producing bipolar chips, giving AMD an advantage since some applications required both types of circuits. Sales were surging, reaching $709.6 million in the four quarters ending in the fall of 1984. Sanders was one of the best-paid executives in the computer industry, earning $948,000 in 1983.

By the early 1980s, the entire United States microchip industry was under pressure from Japanese makers, and in November 1982 Sanders removed himself from day-to-day operations at AMD, partly to concentrate on lobbying in Washington through the Semiconductor Industry Association, of which he was a co-founder. He became a leading voice for the industry, calling for the United States government to get tough with Japanese companies for their allegedly unfair trade practices.

AMD’s fortunes took a turn for the worse in the mid- and late 1980s as a result of changes in the semiconductor and computer industries. In mid-decade, the products of semiconductor start-up firms begun in the early 1980s began catching on. While AMD and other older firms had been offering standardized products, new firms offered semi-finished chips that industrial customers could customize. AMD also fell behind in a new process technology called CMOS, which allowed for faster chips that used less power. The company’s difficulties were further exacerbated when a group of engineers left AMD in 1982 to form Cypress Semiconductor, which specialized in CMOS chips. Other problems were caused by competition from Japanese companies, high production costs, and slumps in the computer industry.

AMD fought back, upgrading its manufacturing plants and moving most production to Texas, where production was less expensive. To cut costs it also put most of its 8,000 U.S. workers on a four-day work week for six months. These cuts were not enough however, and the firm’s no-layoff policy was rescinded in 1986, when 920 people were laid off and research and development spending was cut. AMD also stopped making dynamic random access memory chips, which had not been profitable.

In 1987, AMD acquired chipmaker Monolithic Memories in a stock trade valued at $425 million. Sales for the year reached a record $997 million, but the firm lost $48 million due to costs of the acquisition and the restructuring of its manufacturing plants, which had cost $790 million over a four-year period. To compensate for these costs, AMD raised $172.5 million in 1987 through a preferred stock offering, thus lowering its long-term debt to $136 million.

Despite these cost-cutting measures, a slump in the PC market kept pressure on AMD. The firm was forced to close a plant and lay off another 2,400 workers in late 1988. Partly to deal with these growing problems, AMD restructured its top management in May 1988, creating an office of chief executive. Semiconductor sales remained slow, lowering price margins and causing a four-day shutdown of United States production in September. Profits for the year were only $19.3 million on sales of $1.126 billion. The weakening United States economy led to even poorer results in 1990 as auto makers and the military held back on chip purchases and prices dropped. AMD lost $53.5 million on sales of $1.059 million for the year.

AMD’s recovery was threatened by Intel’s decision to stop licensing second-source rights when it came out with its next-generation computer chip, the 80386. Mostly because its version was faster, AMD had captured 52 percent of the market for the 286 chip, compared with Intel’s 33 percent, and Intel sought to reverse this loss of market share. AMD took Intel to court, claiming that the firm’s previous agreements entitled it to second-source the chip. While the case was in arbitration, AMD set about cloning the chip through reverse engineering so it would have something to fall back on if it lost the legal battle. A secret, windowless lab was set up where AMD technicians shaved away thin layers of an Intel chip and took photographs of it under a microscope. The technicians then copied the chip’s 275,000 transistors, improved the way they were linked, and added 6,000 more transistors. The task took two years, but by August 1990 AMD had constructed a version of the 386 that processed signals 40 million times a second, compared with 33 million for the Intel chip, and used less power.

The AM386 was introduced in March 1991, and captured 15 percent of the 386 market within six months. Its lower power consumption made it popular in the rapidly growing market for battery-powered laptop computers. Intel fought back with an advertising campaign aimed at increasing its name recognition among consumers and with another lawsuit, this one claiming that AMD had no right to use the numerals “386” in the product name of its chip.

AMD won the battle to second-source the chip and use the 386 name, but was criticized by industry analysts for being overly dependent on this one product. The firm also had to fight for the 386 market against new clone companies like Chips and Technologies. Meanwhile Intel’s new 486 chip had captured the upper end of the market, and AMD had not yet succeeded in copying the 486’s 1.2 million transistors.

Despite these difficulties, AMD’s share of the 386 market had increased to 30 percent by the end of 1991, leading to the firm’s highest profit margin since 1984. Intel cut its prices in response to the AM386’s popularity, and Compaq, which had announced plans to use the AM386, switched to Intel’s chip. That blow was mitigated when AMD won a legal battle to use Intel’s microcode, the language that controls how a microprocessor operates. In March 1992 AMD maintained position in its intense competition with Intel by releasing new versions of the AM386 with improved speed and lower power consumption. It also announced the release of its first 486 chip.

AMD’s quarterly sales were at record levels, but industry analysts warned that prices for the 386 chip were likely to plummet because of an over supply. The firm had also been forced by its hard times in the late 1980s to cut research and development from 24.9 percent of sales in 1987 to 17.4 percent in 1991. Faced with limited options because of the great expense of developing new chips, AMD signed an agreement for a $700 million joint venture with Fujitsu Ltd., the largest computer maker in Japan. As part of the deal, the two firms were to buy small stakes in each other. AMD and Fujitsu agreed to develop, manufacture, and market flash memories, a new technology in which memory chips retained information even after the power was turned off. It was expected that flash memories would replace heavy, power-greedy hard drives in portable computers and were expected to be one of the fastest-growing sectors of the computer industry during the 1990s.

A number of Japanese and American companies had already entered joint agreements, but the AMD/Fujitsu deal went further, covering marketing and manufacturing rather than just development. AMD was to supply the design expertise, while Fujitsu contributed manufacturing proficiency. AMD was already the second-largest United States manufacturer of flash memories, behind Intel, which had already signed an agreement with Sharp Corp. of Japan. The Fujitsu deal gave AMD its first manufacturing plant in Japan and strengthened the company’s position in an era of increasing computer industry alliances. The move was viewed favorably by many market analysts, who felt that AMD’s future strength required it to decrease its dependence on clones of Intel chips.