The Man Behind the Microchip Page 20

Why stay? The market certainly said to go. In 1958, electronics stocks had risen twice as fast as the Dow Jones industrial average; one year later, their value increased another 50 percent. Across the country, new semiconductor firms were spinning out of old. In Southern California, Microsemiconductors spun out of Pacific Semiconductors. In New York, Silicon Transistor filed suit against an unnamed firm started by its own former employees. Diotran Pacific emerged from Electroglass in Palo Alto, and Melpar sued a group of employees for starting a firm called Scope in Virginia. One estimate holds that by the end of 1961, between 150 and 200 semiconductor operations had emerged from the handful of companies that had existed in the mid-1950s.80

  Indeed, the frenzy for electronics was so great in 1960 and 1961 that new companies could raise capital on terms similar to those enjoyed by established firms. The Securities and Exchange Commission even felt obliged to issue a warning to consumers about “improper practices in the issuance and sale of space-age stocks.” One newspaper reported that brokers were willing to float an issue for almost any company managed by “a bright electrical engineer under 40 years of age.”81

  The people who worked with integrated circuits at Fairchild had an additional motivation for leaving. Moving the Micrologic (integrated circuit) line from R&D to manufacturing—a process that was haltingly progressing at the time of the spinouts in 1961—had revealed a chasm between these two key parts of the company that could easily be exploited. In the broadest terms, a device passed through three distinct stages from concept to product. First, the R&D division determined the feasibility of a certain design, or else discovered a primary effect that might have some commercial value. In the development stage, which is where Last’s team worked, slivers of silicon were doped, etched, and otherwise tinkered with until a successful prototype, and then a small batch of devices, could be produced.

  The transition from development to the third stage, manufacturing, was in some sense the trickiest part of getting a semiconductor device to market. In manufacturing, several thousand devices were produced every week in a cavernous fab manned by dozens of “girls”—quite a different undertaking from the development stage, in which fewer than 1,000 devices were churned out over several weeks’ time in a controlled laboratory setting.

  A strangely mystical aura surrounded the move from development to manufacturing. The fact that a device worked in development did not automatically mean that it could be reliably manufactured in mass quantities. Problems appeared in the fab that simply had not existed in the lab. Sometimes the problems would disappear for no apparent reason, only to suddenly reappear. Solutions that worked one time might not work the next. Many elements of semiconductor manufacturing were so poorly understood that the problems encountered were given colorful names, such as “Purple Plague” and “Red Death.” Scientists routinely referred to “black magic” and “witches’ brew” in describing their process techniques. At Semiconductor, the appearance of one problem after another in transferring a device or technique to manufacturing was so common that, at one point, Moore was happy to tell Noyce, “No new significant problems have arisen, which is a kind of progress.”82

  By 1961, Fairchild Semiconductor’s R&D and manufacturing operations were not just in different buildings, but in different towns: R&D had moved to Palo Alto, and the manufacturing fab was in Mountain View. The tensions between the two sites ran so deep that Moore confided to Noyce that “our transfer procedure is plagued by meetings that must be attended by everyone to protect their positions.” Moore and the R&D team resented these delays because every day that the Micrologic elements were not coming out of manufacturing was another day that the development pilot line was occupied with building Micrologic devices instead of moving on to new work. Manufacturing, of course, felt that they were doing all they could to move things along, and they complained that R&D was unwilling to send anyone to Mountain View to help manufacturing get through process troubles.83

  In later years, Noyce would try to ease the tensions between development and manufacturing. He pulled his staff together to tackle agendas like this one:

  Communication lines—how can these be shortened? Decisions must be made!

  Reaction time—why are we doing so poorly?

  What are problems on transferring new products?

  a. Enough R&D

  b. Enough factory effort84

  But in 1961, before the first exodus from Fairchild Semiconductor, Noyce seems to have done little to bridge the divide between his development and manufacturing groups. Perhaps he considered this gap and its attendant tensions to be a normal part of the semiconductor business. After all, Bell Labs, the granddaddy of the semiconductor industry, had no manufacturing division at all. Manufacturing was done at a separate, though organizationally related, company: Western Electric. Perhaps Noyce thought that Fairchild would outgrow these difficulties when Charlie Sporck settled into his new job at the helm of manufacturing, or perhaps Noyce wanted to let Sporck and Moore fix the problems themselves, or perhaps Noyce was too distracted with affairs in Italy, Syosset, and elsewhere to pay attention to such nitty-gritty conflicts.

  Whatever the reason for the development-manufacturing divide’s persistence, the companies that spun out of Semiconductor used it to their own advantage. If Fairchild could not manage to build the products it developed, then the development teams would launch their own companies to build them—and try to make a bit of money in the process.

  For nearly a year, when every month seemed to bring the announcement of another new spin-off or another key player’s defection, Noyce seemed a bit stunned. Hurt, too, particularly when his fellow founders left. Noyce had occasionally joined Hoerni, Roberts, Last, and Art Rock on weekend climbs or hikes in the Yosemite back country—excursions so physically draining that the men had to help each other out of the car when they got home. Now three of these men had left, with the assistance of the fourth. Nonetheless, when Noyce could push his personal feelings aside, he could appreciate the siren song of wealth and managerial control. He maintained a philosophical attitude and managed a smile and good wishes for anyone who said they were leaving. He also warned them, his tone always friendly, not to recruit too aggressively from the Fairchild ranks. He would rather not take his friends and former employees to court, he said.85

  Noyce’s sadness at his colleagues’ departures was certainly eased by the fantastic performance of Fairchild Semiconductor throughout 1961. In that single year, Semiconductor doubled both its share of the world semiconductor market and the size of its product line. Largely on the strength of Semiconductor’s growth, Fairchild Camera and Instrument boasted record highs in both profits ($6 million) and sales ($101.5 million), its share price soared, and the stock split two for one for the second time in two years. By the end of 1962, sales were up another 10 percent, profits 14 percent, and some 3,000 people worked for the Semiconductor division.86

  6

  A Strange Little Upstart

  Noyce once said that “the job of the manager is an enabling, not a directive job … coaching, and not direction, is the first quality of leadership now. Get the barriers out of the way to let people do the things they do well.” He had adopted this approach when he ran the lab, and as general manager, he continued it. He wrote personal notes to researchers whose work impressed him. He poked his head into employees’ offices to thank them for their work—and said it sincerely enough that one man so complimented compared the experience to “a hundred percent raise.” Many of the elements of Noyce’s managerial style trickled through the ranks. Supervisors tended to give the people who worked for them a job and let them do it themselves, with little guidance and often little systematic follow-through to check on progress.1

  Noyce’s disdain of hierarchy, present since he assumed the reins as general manager in 1959, was thoroughly imprinted upon the company within five years. A visitor in 1964 described the “lack of adornments” in the Fairchild building, the “informal feel to the place [with] of
fices nowhere near as plush as you’d find at a Motorola, where [the CEO’s] office was about four times the size of Noyce’s office.” Noyce would wander through the main Semiconductor building, admiring aloud the family photos that employees had on their desks. He would stop to talk to anyone about anything and knew many details of his employees’ personal lives. Noyce liked to gather a group of informed people in a room, listen to their opinions, and ideally, get a broad acceptance on the next steps before he made a decision. His staff—a small group of managers—met weekly over cocktails at Chez Yvonne, a local restaurant. This was in sharp contrast to meetings in other parts of the Camera and Instrument organization, which were held not to make joint decisions but to pass commands from John Carter down to his subordinates. Although the visitor in 1964 viewed the casual atmosphere at Fairchild Semiconductor as evidence of “a lack of professionalism”—“they seem to be playing at going into this. The professional management type of idea just isn’t there”—for Noyce, with his egalitarian notions, it was a point of pride.2

  The Fairchild company newsletter once listed the comments made so often by company managers that they had become signature taglines. Sporck’s “executive expression” was a bearish harrumph; Bay’s was a distracted mmmhmmm. For Noyce, the comment was, “Well, what’s new and exciting today?” Noyce forever looked to the future and its technical promise. In 1965, for example, he told a gathering of financial analysts that he expected one day to see integrated circuits inside of “portable telephones, personal paging systems, and palm-sized TVs.” Around the same time, he predicted that “the time will come when every industry will have its electronics shop just the way it has its machine shop right now.” At the end of the 1960s, Noyce enjoyed thinking about how he might build thin tiny television screens that could flip down over a pair of glasses for easy viewing. Kennedy’s call for Americans to land on the moon thrilled Noyce, who eagerly talked with anyone who would listen about how solid-state devices would best be deployed in the moon shot. He was a lifelong devotee of space travel who, well after it became commonplace to push beyond the confines of the earth’s atmosphere, regularly erased his own television interviews to make room for video recordings of various liftoffs.3

  Noyce’s focus on the future and innovation appealed to the creative instincts of many Fairchild Semiconductor employees and permeated the company. To be sure, in the mid-1960s, Fairchild Semiconductor was not a typical semiconductor company. Andy Grove, who joined in 1963, once described Fairchild as “a strange little upstart,” a phrase that captures the essence of the organization. Noyce’s unorthodox management style was just one innovation launched at Fairchild. In the lab, informal company policy allowed “PhDs to play with their ‘toys’ [ideas] for about a year” before expecting results. If an idea appealed to a researcher—for whatever reason—he was free to pursue it. This rather loose definition of relevance led Fairchild researchers to develop roughly one-sixth of all major integrated circuit innovations during the technology’s first two decades.4

  Divisions outside of R&D were likewise highly innovative under Noyce’s leadership. In 1964, Semiconductor replaced the industry’s traditional product-based marketing structure with an application-based selling approach. Instead of a diode or transistor sales manager, for example, Fairchild would have an entertainment-consumer market manager or a military market manager. Each product was shipped with a technical manual that was so much more informative and detailed than any competitor’s that the manuals, written by Fairchild engineers, became products themselves. In 1961, Fairchild announced that its semiconductor products would be available not only through distributors (who sold on commission and represented the traditional sales channel) but also via individual representatives who would buy products outright and sell them at a profit. These stocking representatives in effect became a secondary sales force for Semiconductor. In 1967, the company broadcast one of the world’s first “infomercials”—a half-hour “Briefing on Integrated Circuits,” designed to update and entertain engineers with chipper discussions and colorful illustrations of the state of the art in integrated circuits.5

  Ever since Fairchild’s inception, the focus on innovation had led the company to reject most direct government contract work. Of course, Noyce knew that without the government—specifically, the Department of Defense—Fairchild Semiconductor would not exist. In the company’s first two years, direct government purchases accounted for 35 percent of Fairchild Semiconductor’s sales, and well over half of the company’s products eventually found their way into government hands. The multimillion-dollar Minuteman contract for transistors cemented the company’s success, and the vast majority of Fairchild Semiconductor’s other early customers were aerospace firms buying products to use in their own government contract work. In 1960, 80 percent of Fairchild’s transistors went to military uses, and fully 100 percent of the company’s early integrated circuits were used in defense functions as well. The company worked closely with military contractors in designing and building its products. Even by the mid-1960s, when Fairchild Semiconductor assiduously courted the industrial and commercial markets, its products nonetheless could also be found in surveillance radar and transmitters for space vehicles; in Polaris, Minuteman, and Advent missiles; and in the MAGIC airborne inertial guidance computer, as well as the MARTAC missile-control computer.6

  Though Noyce welcomed the government as a customer and appreciated that federal mandates—such as one issued in April 1964, that required all televisions be equipped with UHF tuners, a law that effectively forced the introduction of transistors into every television in the United States—could benefit Fairchild Semiconductor, he believed there was something “almost unethical” about using government contract money to fund R&D projects. “Government funding of R&D has a deadening effect upon the incentives of the people,” he explained to a visitor in 1964. “They know that [their work] is for the government, that it is supported by government dollars, that there is a lot of waste. This is not the way to get creative, innovative work done.” He added, “The best way to get something done is to have enough confidence in yourself and your men to do it yourselves.”7

  Noyce also resented that the proposal requests issued by the government were “written as if everyone who bid was a crook,” complaining to his son, who was only a boy, that “it’s not enough to say they want a certain result; they specify every test to be run, every result they want to see.” The bureaucracy this represented—and required in the bidding company—seemed wasteful to Noyce, and he also worried that such specificity limited a lab’s creative flexibility to explore the “interesting slop” that might unexpectedly emerge in the midst of research. “A young organization, especially in the electronics industry has to be fast moving,” he explained in 1964. “It runs into problems with the unilateral direction mandated by government work.” By this point, the company was relatively well established, and Noyce reminisced, “We were a hard, young, hungry group. [Our attitude was] ‘We don’t give a damn what [money] you have [to offer], buddy. We’re going to do this ourselves.’” Gordon Moore shared Noyce’s beliefs. Consequently, while other firms in the early 1960s used government contracting as the primary source of R&D funding, less than 10 percent of business at Fairchild was contracted directly by the government in 1963. “And we like it that way,” Noyce hastened to tell a reporter.8

  In the early 1960s, Fairchild began to internationalize its business. Before 1965, with only a few exceptions (among them, SGS, the Fairchild-Olivetti-Telettra joint venture), the United States semiconductor industry researched, developed, manufactured, assembled, tested, and marketed every semiconductor from a location on American soil. Fairchild broke the mold. In May 1962, Noyce’s assistant Jerry Levine, who had a penchant for travel and a nose for business, cashed in three years’ accumulated vacation days and paid his own way to Hong Kong because he thought Fairchild should consider building an assembly plant there. Levine had been talking about this for months. La
bor costs were low in Asia—$1 a day was considered a good wage—and the workforce was certainly capable of assembling and packaging low-performance transistors destined for the entertainment market, which did not have the exacting standards of the military or computer markets.

  Nearly everyone who heard Levine’s idea thought he was crazy to imagine that Fairchild could send nearly finished semiconductor devices 5,000 miles across the ocean, teach people from a vastly different culture how to assemble and package them, and then ship the devices back to California for testing. Noyce, on the other hand, did not see why Levine’s idea should not work. American transistor radios had been fabricated in the Philippines and India for years, and the Bay Area electronics firm Ampex had a successful, small, low-skilled assembly operation in Hong Kong. “Space in Hong Kong. I should work out responsibilities for organization of o[ver]seas operation.” Noyce wrote to himself after his meeting with Levine. “My ball,” he added, meaning that he needed to take the next step.9

  Noyce knew that paychecks for people who assembled and tested Fairchild’s products were the single largest source of the company’s expenses. Texas Instruments and Motorola had both automated chunks of their low-skilled test and assembly processes to cut costs, but a brief experiment with a “semi-automatic assembly line” at a Fairchild transistor plant in 1960 had convinced both Noyce and manufacturing head Sporck that the upfront costs of automation were daunting, and the payoff uncertain since production processes changed so quickly that machines installed one year might need to be retooled the next. One month after Levine’s initial trip to Hong Kong, Noyce asked Sporck and Julius Blank, who served as facilities manager, to visit Hong Kong themselves. Sporck and Blank returned to California convinced that Levine’s idea was worth trying. Hong Kong offered the three key features of a good plant location: low wages, low building costs, and a large potential workforce of women whose small fingers and dexterity were considered critical for assembly work. Adding to the attraction were special tariff regulations allowing companies that shipped components overseas for assembly to re-import them into the United States for testing and distribution under a special “low-value-added” duty.10