The drive for digital electronics that spelled Fairchild Camera and Instrument’s loss was Fairchild Semiconductor’s gain. In the past year, the air force had begun shifting its major contractors away from systems based on analog computers—which suffered from easily blown vacuum tubes and a multiplicity of moving parts—towards all-digital systems, which were faster and more accurate than their analog counterparts. The air force had further begun requiring the use of silicon transistors (which, unlike germanium devices, could withstand high temperatures and avionic jostling) in the digital computers for its missiles and airplanes.31
Shortly after Camera and Instrument lost the B-58 camera contract, Fairchild Semiconductor marketing manager Tom Bay came across an article detailing the difficulties that IBM’s Federal Systems Division faced in its efforts to build a navigational computer for the B-70, a long-range strategic bomber nicknamed “the manned missile.” When he read that IBM’s most pressing problem was its lack of a silicon transistor for the computer, Bay, who was familiar with guidance systems, thought immediately of Sherman Fairchild’s connection to IBM. Sensing opportunity, Bay and Hodgson convinced Sherman Fairchild to arrange for Bay and Noyce to meet with the IBM engineers working on the B-70 computer in Owego, New York.
IBM wanted a transistor that could withstand high temperatures and that could switch quickly. Bay recalls the specs called for a device that could switch 150 milliamps with 60-volt capability at 50 megacycles—faster than any silicon transistor then on the market and faster than many germanium devices, as well. Moreover, IBM wanted 100 of them. Noyce listened intently to the engineers and then said simply, “Sure. We can do that.” Noyce’s confidence—or was it bluffing?—both impressed and surprised Tom Bay, who at the time noted that “Bob is so articulate, no one questions [him].” Their three-month-old company had yet to build even a single basic transistor, and here was Noyce coolly promising 100 state-of-the-art devices, with “never a doubt in his mind that we could do it,” as far as Bay could see. Perhaps Noyce counted the fact that Fairchild Semiconductor had not built any other transistors a benefit—no established standards, practices, equipment, or training meant nothing to undo or retool for the IBM device. They could, in effect, build the company around IBM’s needs.32
Noyce’s assurances did not overcome the IBM engineers’ immediate doubts about Fairchild Semiconductor, a complete unknown distinguished in the industry only by the rebellious history of its founders. Although IBM had no real alternative supplier to Fairchild (the best approximation of the device they needed, a Texas Instruments component, had failed in testing), it took a private meeting among Sherman Fairchild, Dick Hodgson, and IBM chief Thomas Watson, Jr.—a meeting with a singular theme: your largest shareholder has invested more than $1 million in these men, so you should trust them—to persuade the IBM Federal Systems engineers to take a chance on Fairchild Semiconductor. By February 1958, the young men had an order for 100 silicon transistors. IBM agreed to pay $150 for each transistor, this at a time when basic germanium devices were selling for less than $5. Fairchild Semiconductor was in business.33
THE IBM ORDER made Fairchild Semiconductor. IBM left little to chance, carefully specifying not only the device’s electrical parameters, but also the packaging and testing procedures the young company should use in manufacturing the transistors. Noyce and the other scientists at Fairchild Semiconductor agreed they could only achieve the sort of reliability and speed that IBM wanted if they built double-diffused silicon transistors: devices, that is, built out of silicon and with two P-N junctions. They would build mesa transistors like the ones they had worked on at Shockley Semiconductor. The only question was whether the transistors should be diffused PNP or NPN. The group decided to split the company in two and try to build both devices, with Gordon Moore heading the NPN effort and Jean Hoerni leading the drive on a PNP device.34
With semiconductor manufacturing still in its infancy, the Fairchild Semiconductor men “had to develop our own equipment as we developed the processes,” as Gordon Moore put it. They had done this type of work at Shockley and could quickly assume responsibility for their areas of expertise. Sheldon Roberts, section head for materials processing and metallurgy, took charge of growing the silicon ingots, slicing them into “wafers” approximately the size of a dime, and polishing them until they gleamed like mirrors. Ultimately, dozens of transistors would be etched onto each wafer, but first the silicon needed to be diffused—doped with impurities in high-temperature furnaces so that some areas of the silicon wafer were P-type and others N-type. Gordon Moore and Jean Hoerni oversaw the diffusion process, with Hoerni taking charge of the theory (determining how long to diffuse the wafers and at what temperature) and Moore overseeing the practical necessities of building the furnaces. Moore’s needs were so specialized, and the pickings so thin, that he ultimately had to order the elements he needed from a company in Sweden and design and build the furnaces himself.35
Once the silicon was properly diffused, it was time to start differentiating the individual transistors. This was done through a process called photolithography, an area that Noyce and Jay Last led together at Fairchild. The two men created a pattern that showed where every transistor would appear on the wafer, how the current would pass through the transistors, and where the transistors would be attached to the canisters that would then be plugged into the IBM system. This pattern would then be shrunk hundreds of times in succession until it was so small that multiple copies could be lined up side by side on a small glass plate called a mask. When Noyce needed to build the camera that would reduce the pattern, he went to a photography store in San Francisco, where he rummaged through a bin of 16-mm movie camera lenses until he found three that, while not flawless, could be aligned in such a way that the errors on them did not affect the process of shrinking the pattern. This cobbled-together machine became the prototype of the step-and-repeat cameras used throughout the industry.36
To transfer the patterns on the mask to the silicon, Noyce and Last needed to coat the surface of the silicon with a light-sensitive resin. Eastman Kodak had developed a resin (for use in printed circuit boards) whose chemical composition Noyce and Last could modify to meet their needs. After this resin, called “photoresist,” was applied to the wafer, the mask was placed on the coated wafer. Then the wafer and mask were exposed to light, acid, and a dopant that added one more P-N junction to the surface of the silicon wafer while also etching away areas that had not been covered by the photoresist. These processes happened over the surface of the entire wafer, so that all the transistors were processed simultaneously, rather than one at a time.37
Once the silicon wafers emerged from the photoresist process, the transistors needed to be cut apart and tested. Not every device that started production worked by the end. Out of 100 transistors on a wafer, for example, the number of working devices (called a “yield”) would be somewhere between 10 and 50. Yields were low because contamination of any sort ruined the electrical characteristics of the device—and contamination was rampant, despite plastic sheeting and other rudimentary attempts to keep the lab clean. Moreover, even a slight change in the moisture in the room or a bit too much dust could cause a device to fail.38
Victor Grinich, who also helped to define product applications and evaluation protocols for new devices, took charge of testing. Julius Blank designed the manufacturing facililty (called the “fab”) and ran plant engineering in collaboration with Eugene Kleiner, who was also responsible for general administration since he had known a banker. Blank and Kleiner also supervised the women (invariably called “girls”) who cut the transistors apart, wired them into canisters, and packaged the finished product for shipment to customers. Women were hired for these jobs because it was believed that their small hands and well-developed fine-motor skills would ideally suit work with tiny devices and small wires—and they could be paid less than men.39
For the founders, it was a very egalitarian arrangement. “Noyce was the tech
nical head of the lab, and that was it for organizational structure,” according to one founder. “The rest of us were pretty much on equal footing. Everybody wore as many hats as possible.”40
The group began quietly recruiting from Shockley Semiconductor Labs. Capable Shockley employees knew that if they read the Palo Alto Times carefully, they would at some point see their current job description listed in the classified advertisements. “Fairchild Semiconductor did everything but put our names on the ads,” recalls one employee. Soon Harry Sello, who worked at Shockley, joined, as did Dave Allison, who came to work on the NPN transistor, and C. T. Sah, a gifted physicist. A small contingent of technicians also came to Fairchild from Shockley.41
A general manager joined the company in February 1958. Ed Baldwin was a former paratrooper who had managed the diode operations at Hughes, a major military subcontractor and one of the nation’s leading producers of silicon semiconductors. In addition to his experience in precisely the markets Fairchild Semiconductor targeted, Baldwin had a PhD in Physics and a presence as compact and sturdy as a cannon. The eight founders offered him a stake in the company equivalent to their own—$500 for 100 shares—but he did not accept. Privately Baldwin was lobbying Hodgson for stock, arguing that his managerial position justified his having a larger stake in the company than did the founders. Soon some dozen engineers from Hughes, most of them specialists in manufacturing—the eight founders’ weak point—were on the Fairchild Semiconductor payroll.42
Baldwin’s job was not easy. After an initial $50,000 loan, Camera and Instrument transferred funds to Semiconductor only to reimburse properly documented expenses, in effect keeping Baldwin and his team on an allowance. Although this did provide a workable method of cost accounting for the parent firm, it also created a burdensome layer of bureaucracy for Fairchild Semiconductor.43
Baldwin’s slogan could have been “Think Big and Focus.” He told the eight founders that they needed to develop an organizational chart. They needed to begin planning—immediately—for a much bigger manufacturing facility, even though they had yet to build a product. They should separate engineering from manufacturing, and both of them from Research and Development. This was not a radical suggestion; inasmuch as there were traditional operating procedures for an industry as young as semiconductors, these sorts of divisions would qualify. Baldwin did not particularly admire the founders’ abilities to perform multiple jobs. “Do one thing and do it well,” he told Noyce. Baldwin wanted an instrumentation expert to build the test equipment and the preproduction engineering to be overseen by someone who had actually put products into production in the past. When Tom Bay told him that he thought Fairchild Semiconductor might have $15 million in revenues in five years, Baldwin ordered him to “shoot ten times that high.”44
Despite his contributions, Baldwin was always outside the founding group. The rapport the eight of them shared was dynamic and impermeable. They worked together ten or twelve hours each day, not counting the trips to Rupert’s Bar, where they liked to go for drinks in the evening. They often found themselves standing in a circle when they were together, their shoulders nearly touching, each man holding one conversation with the man on his left and a different one with the man on his right (and perhaps a third with someone across the circle). Noyce loved these moments, loved the buzz of talk and the smell of the cigarettes many of them held between their lips. If he reached in his pocket for a smoke and discovered an empty pack, he would crumple the wrapper, toss it on the ground, grab a cigarette out of his neighbor’s pocket (without asking and almost without looking), pound it on his leg, and pop it in his mouth so the guy from whom he took it could light it. Were it filtered, Noyce would grumble something about sissies. All the while, he and the other fellows would maintain their end of two or three conversations.45
A formal photo from the founding period hints at this rapport. The eight founders sit around a table under a flower-print, fringed umbrella. They are seated in a circle with Noyce, as always, front and center. They all wear suits—most of them owned only one—and ties, which give them a serious air, as do the papers and books strewn across the table. But every man wears a smile big enough to be called a grin. They are clearly having the time of their lives.
By May of 1958, the NPN transistor Moore’s team had built for IBM was ready to move into production. By early summer, Fairchild Semiconductor had delivered its promised 100 devices to Owego, New York. Though he was not one for sentiment, Noyce kept the check stub from the transaction for the rest of his life. In August, Fairchild Semiconductor brought its NPN transistor to Wescon, the six-year-old trade show sponsored by the West Coast Electronics Manufacturers Association. There the company’s founders, several of whom presented papers together, were elated to learn that theirs was the only double-diffused silicon transistor available on the open market. “We scooped the industry!” Noyce whooped to a group of Fairchild employees a few days after the show ended. “Nobody [is] ready to put something like this on the market … [and there is] no prospect of anybody getting in our way in the immediate future.” Indeed, Fairchild kept its monopoly on the device for over a year.46
5
Invention
The infancy of Fairchild Semiconductor undoubtedly ranks as the most intellectually fertile time of Noyce’s life. Seven of his 17 patents, including his most important, for the integrated circuit, date from the 18 months after the company was launched, when Noyce served as the director of R&D and oversaw Fairchild’s research efforts. During this time, he focused as much attention as possible on his own scientific work. He spoke at technical conferences on topics such as “Switching Time Calculations for Diffused Base Transistors.” His mind was so filled with ideas that sometimes he would rise in the middle of the night to write them down while Betty and the children slept. In the office, he kept careful lab notebooks in which he adopted a highly didactic tone—“let us look at the following structure,” “we may set these criteria”—almost as if he were lecturing to himself in page after page of notes, charts, figures, and oscilloscope readings. Sometimes he scrawled complex mathematical equations over entire pages; at other points he slowed down long enough to draw up a “plan of calculation” before launching into the math.1
Truth be told, though, the brutally methodical labor of science—the careful working and recording of one’s way through experiment after experiment, each one only slightly different from the iteration that preceded it—interested Noyce far less than the moments when a new idea came to him. At his core, Noyce was an almost compulsive idea generator, a mental perpetual motion machine. Thomas Edison famously declared genius to be 99 percent perspiration and 1 percent inspiration, but Noyce preferred to spend as much time as possible in the inspiration stage.
Some scientists do their creative work by starting small and building up. When William Shockley wanted to invent, for example, he liked to pull out every publication he could find in the relevant fields. He would then spread the patents and papers across his desk and try to make novel connections among them or to identify potentially fruitful areas that had not yet been investigated by other researchers.2
This assembly line inventiveness was not for Noyce. Unlike Shockley, he never sat down and told himself he needed to invent something before he could stand up again. His approach, Noyce once told a friend, was to know the science cold and then “forget about it.” He did not slog or grind his way to ideas; he felt they just came to him. When he heard Picasso’s famous line about artistic creativity—“I do not seek; I find”—Noyce said that he invented in the same way.3
Noyce’s creativity often required a kick start, usually in the form of a practical question from a colleague. Once his attention was engaged, Noyce did not start small, and he did not turn to journals or patent files for ideas. Instead, he tried to “think about the fundamentals of the physics”—as big a starting point as possible—and he refused to ask himself whether or not an idea ought to work according to the most current resea
rch in the field. In his opinion, there were only two relevant questions in the earliest stages of scientific innovation: “Why won’t this work?” and “What fundamental laws will it violate?” If an idea seemed within the realm of physical possibility, then Noyce deemed it worthy of exploration—conventional wisdom on the topic be damned.4
Noyce’s try-anything approach, by no means unusual among technically or mathematically gifted thinkers, meant he consistently generated ideas that seemed implausible, more flashes of intuition than products of careful science. Often these ideas proved dead ends, but occasionally they were brilliant. No one knew this better than Gordon Moore, Noyce’s second in command at the lab, and a deft screener of what he called “Bob’s many ideas, some of them good.” Moore excelled at the perspiration work of science, which made him an ideal creative complement to Noyce.5
One of the most memorable examples of Noyce’s inspiration working thanks to Moore’s perspiration came in 1958, when Moore, who was working on the IBM transistor, was searching for a way to use a single metal to make contacts to both the P- and N-type silicon at the surface of the transistor. (Contacts were little dots of metal to which the wires that connected the transistor to its canister were attached.) Western Electric had experimented with using two different metals for contacts—aluminum on the P-type silicon and silver on the N-type—but their process was so complicated, and their yields so dismal, that Moore wanted something better for Fairchild.
The Man Behind the Microchip Page 15