Fairchild Semiconductor
A Spectacular Rise and Gradual Fall
This is part two of a multipart series on the birth of the silicon semiconductor industry in the USA. You may be interested in part one, which covered Shockley Semiconductor Laboratory.
On the 18th of September in 1957, seven of the traitorous eight (Gordon Moore, Sheldon Roberts, Jean Hoerni, Julius Blank, Victor Grinich, Eugene Kleiner, Jay Last, Robert Noyce) left Shockley Semiconductor Laboratory. On the 19th of September in 1957, those eight men signed the contract that created Fairchild Semiconductor Corporation, with Fairchild Camera and Instrument as the funder and parent, at 844 South Charleston Road.
Getting to that point was difficult. First, there weren’t really any venture capitalists on the west coast of the USA at the time, and Noyce who was the Director of Research and Development at Shockley Semiconductor wasn’t fully on-board. The group of seven wrote a letter to Hayden, Stone, and Company where Kleiner’s father had several contacts. The letter was circulated, and it was one Arthur Rock who found it interesting. He was struck by the fact that these seven were handpicked by Bill Shockley who’d had a rather free hand in picking talent, and who was known to Rock (and to many) by reputation for having had talent in picking talent. Rock did have a serious concern. The group needed a single leader with knowledge and talent in order to attract other investors. The group of seven stated that they did, indeed, have a leader, Robert Noyce. The issue was that he felt some obligation to Shockley.
Robert Norton Noyce was born on the 12th of December in 1927 in Burlington, Iowa to Ralph Brewster Noyce and Harriet May Noyce (née Norton). His father was a minister, and his mother stayed at home and been by a minister. Both of Noyce’s parents were college graduates. He had three brothers growing up in Grinnell, Iowa, and of the four, Robert was the more dramatic. As a young man, he liked model airplanes. When these became a bit worn and drab, he’d light them on fire and toss them from the attic window. He also fashioned his own radio and controls for a plane, and then grabbed a motor from an old washing machine, attached a propeller and flew it from the roof of Grinnell College. At Grinnell (and later MIT), he was a singer, actor, oboe player, and accomplished swimmer. He graduated with his BA in physics and mathematics in 1949. He received his doctorate in physics from MIT in 1953. His doctoral dissertation was titled “Photoelectronic Study of Surface States on Insulators,” he had worked at the semiconductor research group at Philco for three years, and after that he went to work for Shockley. He had confidence, technical knowledge, and charisma. At Shockley Labs, he led the senior technical staff, recruited folks, set salaries, and acted as a buffer between Shockley and the company’s employees under his management. Sheldon Roberts reached out to Noyce and filled him in on the conversation with Rock, and he asked Noyce to reconsider. While the call went through the night, Noyce agreed to meet with Rock and Rock’s boss Alfred “Bud” Coyle along with the seven the next morning at the Cilft Hotel in San Francisco. Roberts collected the traitorous eight in his family’s station wagon and they rode shoulder to shoulder from Mountain View to the hotel. At the meeting, Noyce went from being hesitant and obstinate to being passionate and loquacious. After he spoke, none of the rest of the group had much to add. He advocated for the use of silicon substrate, admitted that manufacturing cost would be the main issue, and then laid out the idea that devices could one day be made so cheaply that they’d effectively be disposable.
It was Bud Coyle who made the biggest difference in what was to come. He mentioned the group to Sherman Mills Fairchild. Fairchild was a well groomed and stylish millionaire in his sixties who liked women, parties, architecture, cooking, jazz, dance, philosophy, tennis, and making more money. He was a man who’d never married, never had children, and worked relentlessly. He’d inherited his father’s estate (which included the largest individual holding of IBM) on the 31st of December in 1924, and with that wealth, he founded over seventy companies. He’d have been a successful man either way having achieved much on his own. Despite suffering through tuberculosis and dropping out of college, he invented the synchronized camera shutter and flash. He later invented an aerial camera with a high speed shutter inside the lens that reduced image distortion. This won him a contract with the US government during the first world war, but it came a little late. Undeterred, Fairchild developed a more advanced camera, founded Fairchild Aerial Camera Corporation in 1920, and sold several aerial camera systems to the US Army. This business did quite well during the second world war with Fairchild supplying more than ninety percent of the aerial cameras used by the Allies. This company became Fairchild Camera and Instrument, and the president, John Carter, was tasked with talking with the traitorous eight by Fairchild. Carter, along with executive vice president, Richard Hodgson, flew to California to meet with the eight, and they invited Kleiner and Noyce to make a visit to Syosset, New York. They met at the Fairchild Camera headquarters, and they also met in the home of Sherman Fairchild. At this meeting, the eight were named the California group and somewhat informal but still… official negotiations began.
So, two bankers, Sherman Fairchild, and the California group were going to start a company to make and to sell transistors, a new market that had just reached about $7 million in sales. At their meeting on the 18th of September in 1957, Rock pulled out ten $1 bills and each of the eight and the two bankers wrote their names ten times. These were the contracts with each other. Each of the men would own 100 shares of Fairchild Semiconductor Corporation, Hayden, Stone, and Company would own 225 shares, and 300 shares were held in reserve for managers to be hired at a later date. With Fairchild, things were slightly more formal and more complicated. They reached an agreement for $1.4 million paid out over the first 18 months of operation and more money contingent on progress. That would be roughly $16 million in 2024 dollars. This money was provided under the condition that Fairchild Camera and Instrument had the majority vote through a trust and could exercise an option to buy all of the stock of the company for $3 million at any time. After that point, but within seven years, that buy out price would rise to $5 million so long as the company posted profits for three consecutive years. Additionally, each of member of the California group contributed $500 of their own money, and Richard Hodgson provided $3000 to Noyce to cover the new company’s operating expenses until formal organization was completed. On the company’s documents, Hodgson served as the chairman but remained in New York. Noyce was the director of research and development, and he also had a board seat. Everyone involved, and Hodgson most of all, asked Noyce to serve as the company’s general manager, but Noyce declined. His stated reason was his own lack of managerial experience.
All of the California group resigned from Fairchild on the 18th of September in 1957. At least Moore and Noyce were upset by this, but I can imagine most of the group had similar misgivings. They’d developed the technology they were about to produce and sell largely in secret at Shockley Labs. They were about to build silicon transistors with diffusion and thin-film that they’d only learned by being at Shockley. Despite their own uneasiness, and largely due to Bill Shockley’s unique ability to repulse people, they formed Fairchild Semiconductor Corporation at 844 South Charleston Road, just twelve blocks from Shockley, on the 19th of September in 1957.
In a span of just five months, Fairchild Semiconductor had built out their R&D lab in Palo Alto: Roberts developed the process and techniques for growing silicon crystals, Last and Noyce developed photolithographic masking techniques using three 16mm lenses from movie cameras, Moore developed diffusion methods and techniques for making electrical contacts with aluminum, and Julius Blank, Victor Grinich, and Eugene Kleiner built manufacturing and test equipment. The company’s first sale and delivery was to IBM for 100 transistors priced at $150 per transistor. This was the first sale of a double-diffused silicon transistor (2N696/697 NPN), and this new product was shown to the public at the Wescon electronic components trade show in August of 1958. Early in 1959, Noyce became the general manager at Fairchild Semiconductor.
The company’s next major challenge was reliability within the Minuteman missile’s guidance and control system. With the kind of physical forces experienced in a ballistic missile, tiny bits of metal flaking or coming loose inside the transistor’s package could short exposed junctions on the surface of the silicon. The solution was something that Hoerni had already been contemplating. As early as the 1st of December in 1957, he’d thought about adding an oxide layer on the top of a silicon wafer to protect junctions between p-type and n-type semiconductor regions. Coupled with Kurt Lehovec’s p-n junction isolation innovation known as a guard ring, this evolved into the planar process (as the finished product had a flat surface). Hoerni patented this process in January of 1959 and had a working planar transistor in March. This functioned as he had predicted, and Fairchild won the contract for Minuteman. By September, the company had sold $6,500,000 dollars worth of planar silicon transistors, and Fairchild Camera and Instrument exercised their option to buy out Semiconductor, and the traitorous eight earned $300,000 each (about $3.25 million in 2024 dollars).
Planar transistors made significant gains in the commercial market starting in April of 1960 with the 2N1613 planar NPN. The cost to produce a single transistor had fallen to just $0.13 and each one sold for about $1.50 giving the company a 91% margin on each unit sold. In March of 1961, the company succeeded in epitaxial deposition on silicon. This allowed layers of crystal to have different doping characteristics from the silicon substrate, which then yielded speed increases and voltage requirement decreases (at this point, I really wish I could recall more of physics courses). This advancement earned the company a contract from Seymour Cray at Control Data for the development of a transistor that could switch in less than 3ns. This requirement was met by Hoerni with the use of gold-doping and epitaxial deposition by the 2N709 NPN in July of 1961. Control Data then ordered a bit more than 10 million 2N709s signaling to the world that the era of germanium was ending. By the close of 1961, Fairchild Semiconductor was the largest producer of high-performance silicon transistors in the USA, and they enjoyed a near monopoly for transistors in high-reliability applications.
Of course, the planar process had a much larger impact than simply creating more reliable transistors. While that wasn’t a small advancement, Noyce had far larger ideas. He wanted to create a a truly monolithic integrated circuit in which the diodes, transistors, resistors, and capacitors were diffused into the silicon wafer and connected via the deposition of aluminum lines directly onto the top of the oxide coating. A patent for this was filed by Noyce in July of 1959. Jay Last was then tasked with creating a proof-of-concept, and he assembled a team including Sam Fok, Isy Haas, Lionel Kattner, and James Nall to develop an IC on the basis of Hoerni’s and Noyce’s patents. Robert Norman then designed a resistor transistor logic circuit. For the next year, Last and his group worked on fabricating the device. On the 26th of May in 1960, their IC had four transistors and five resistors in a flip-flop configuration and was tested successfully. The final commercial version of this was brought to the world in a press conference at the Institute of Radio Engineers Show in New York in March of 1961. By the end of the year, the company had shipped half a million dollar’s worth of ICs with a considerable back log of orders. Their early ICs were things like the 900 (a buffer), the 901 (a three gate counter adapter), the 902 (flip-flop), and the 959 (a 4-bit latch). The Micrologic product line eventually included 25 different standard RTL ICs. The industry responded rather quickly with General Motors’ airborne MAGIC I being built from Fairchild ICs and made operational in late 1962. NASA’s Apollo Guidance Computer built by Raytheon used 200,000 ICs.
In 1963, Chih-Tang Sah and Frank Wanlass developed CMOS MOSFET logic. The same year, Robert Widlar and Dave Talbert developed the op-amp. In 1965, the company opened a semiconductor assembly plant in the Navajo Nation in Shiprock, New Mexico. This location primarily manufactured DTL chips, op-amps, and custom ICs for mainframes. Fairchild was the dominant player in the DTL market, and the company’s op-amps were likewise dominant in the market (the μA741 of 1968 was the most popular IC op-amp ever made).
Profits for the company in 1961 passed $3.8 million, and then passed $8 million by 1965 (over $80 million in 2024 dollars). These numbers reflect the wider market at this time. More mature devices could be sold for nickel allowing for less expensive computers to be made, while newer devices with far higher prices were used in the more expensive and more powerful machines used by the military, the government, and research firms/institutions. For all of this success, Noyce was made the corporate vice-president, and he still maintained control of the semiconductor division. But it’s also in the mid-1960s that the first issues began to appear. The company had grown so quickly and had so many different physical locations that the transfer from R&D to manufacturing began to take more and more time, and more and more meetings. As the company became more bureaucratic, internecine political fighting became more common. This was especially true in the sense of the founders vs Sherman Fairchild’s cohort. This disorganized, bureaucratic, politically fraught company began missing delivery deadlines. For one quarter, the company actually only fulfilled a third of its orders. While Fairchild’s DTL had been a chief money maker, it was largely replaced in the market by Texas Instrument’s TTL technology which was far faster. In the autumn of 1967, the company posted its first loss, $7.6 million. In October, the board chose to sell of all of Fairchild’s unprofitable divisions. The president of the company, John Carter, resigned. The board then passed on Noyce as the replacement for Carter, choosing Richard Hodgson instead. A few months later, Hodgson was told he’d report to an office of the chief executive which consisted of Noyce, Walter Burke, Joseph Wharton and Sherman Fairchild while the board searched for a permanent CEO. Charlie Sporck who’d been head of semiconductor manufacturing and then general manager following Noyce’s promotion left to become CEO at National Semiconductor, and 35 other employees followed him. Passed over for promotion, losing employees left and right, constantly engaged in corporate politics, and forced to function reactively, Noyce then began planning a new venture with Gordon Moore, Intel. Noyce and Moore were not alone in their departure, and among the people they recruited was Hodgson. While Fairchild’s fairchildren are often looked to as a unique occurrence, this wasn’t unique. While it was certainly more common at Fairchild, other semiconductor companies had the same problem. There was fierce competition for brightest minds at this time, and the number of companies in the industry was rapidly increasing. To replace personnel, Sherman Fairchild hired Lester Hogan as CEO (at the recommendation of Noyce) who then began hiring folks from Motorola's semiconductor division from which he himself hailed. Noyce then formally resigned from Fairchild on the 25th of June in 1968.
Hogan’s choice to recruit folks from Motorola earned Fairchild a lawsuit. It was a large case with 75 witnesses, 560 evidentiary exhibits by Motorola, and 850 exhibits by Fairchild. This suit was decided in 1973 in favor of Fairchild, but the win wasn’t very sweet. The judge, William Copple, stated that Motorola had lost nothing other than employees which is the normal cost of doing business as Fairchild really didn’t gain any competitive advantage as shown by their continued poor market performance.
In 1973, the company developed the first commercial charge-coupled device as an image sensor. This effort was led by Gil Amelio who’d first developed the technology while at Bell Labs before being hired by Fairchild. It should have been an immediate hit, but the timing was poor. The US entered a rather bad recession the same year.
In 1974, Fairchild announced a new 8-bit CPU, the F8. This had been developed David Chung who’d previously been the leader of the processor division at General Instrument with much of the design and layout done by Robert F. Hartmann. Following the announcement of the CPU, General Instrument filed a lawsuit claiming that Fairchild had stolen trade secrets from them. Despite the on-going lawsuit, Fairchild released the 8-bit 3850 CPU and 3851 PSU (program storage unit) which together comprised the F8 system in its minimal configuration in the autumn of 1975. It was made with an NMOS process and utilized a +5 V and +12 V power supply. In addition to being the CPU, the 3850 contained 64 bytes of RAM and two 8-bit bidirectional I/O ports. The 3851 had 1K ROM and handled instruction decoding. These two chips could be optionally accompanied by either the 3852 DMI (dynamic memory interface) or the 3853 SMI (static memory interface) allowing the F8 system to control more RAM. The final optional component would by the 3854 DMA (directory memory access) controller were a customer also utilizing the 3852. The 3850 was shipped in a 40-pin DIP and could operate at 1MHz to 2MHz. The legality of the CPU was largely settled in February of 1976 when Fairchild reached a cross-licensing agreement with Olympia Werke (this was the company from which General Instrument had licensed their CPU technology). The processor that caused all of this ruckus wasn’t even the basis for the F8 design, but it was a four chip system. Despite this, the court case continued into the 1980s without sales of the F8 having been impacted. The F8 sold very well. Most people aren’t familiar with it as it was mostly used in embedded applications as a two chip microcontoller.
With the F8 CPU, Fairchild launched the first home video game console to use ROM cartridges, the Channel F. This system also pioneered the razor and blades pricing model where the console was sold at a loss to attract cartridge sales where the real money would be made. It had 64 bytes of RAM and a 2K video buffer. It could output 104 by 60 pixels with eight colors, and it had a three tone beeper. The system retailed at $169.95 upon its market release in November of 1976, and it sold around 350,000 units. After the release of the Atari VCS, the system was moribund and Fairchild sold it to Zircon in 1979.
Fairchild’s fortunes were fading, and the company turned to the one market where they’d always seen success, hardened ICs and isoplanar ECL circuits. Still, they were operating in those markets at a loss, and what profits they did make originated with patent licensing. The company was sold to National Semiconductor for $200 million in 1987, came back to life in 1997 with a sale to management for $550 million (this was really a new company with the old name and the old patent portfolio though), and it was then sold to ON Semiconductor (the spinoff of Motorola’s semiconductor products division) in 2016 for $2.4 billion.
The most lasting legacy of Fairchild, aside from groundbreaking technological advancements, is the list of companies it spawned. Among the most prominent are the following:
Amelco, 1961: Jean Hoerni, Eugene Kleiner, Jay Last, Sheldon Roberts
Intel, 1968: Gordon Moore, Robert Noyce
Intersil, 1967: Jean Hoerni
Kleiner, Perkins, Caufield & Byers, 1972: Eugene Kleiner, Tom Perkins, Frank J. Caufield, Brook Byers
Signetics, 1961: David Allison, David James, Lionel Kattner, and Mark Weissenstern
Advanced Micro Devices, 1969: Jerry Sanders, Ed Turney, John Carey, Sven Simonsen, Jack Gifford, Frank Botte, Jim Giles, Larry Stenger
Rheem Semiconductor, 1959: Ed Baldwin
Electroglas, 1960: Art Lasch
Sequoia Capital Management, 1972: Don Valentine
Altera, 1983: Robert F Hartman, Robert Smith, James Sansbury, Paul Newhagen
LSI Logic, 1980: Wilfred Corrigan, Bill O’Meara, Rob Walker, Mitchell "Mick" Bohn
National Semiconductor, 1967 (restart): Charles Sporck, Floyd Kvamme, Pierre Lamond, Donald Valentine
Apple, 1976: Steve Jobs, Ron Wayne, Steve Wozniak (Mike Markkula as chairman with funding from Sequoia and Arthur Rock)
VLSI Technology, 1979: Jack Balletto, Dan Floyd, Gunnar Wetlesen, Douglas Fairbairn
Of course, from these companies many more companies formed. With two venture capital firms, some descend by investments and board seats. In other cases, veterans of the fairchildren started fairgrandchildren like NVIDIA, SanDisk, and eBay.
I have readers from many of the companies whose history I cover, and many of you were present for time periods I cover. A few of you are mentioned by name in my articles. All corrections to the record are welcome; feel free to leave a comment.