Noyce, Kilby, and the Integrated Circuit

In July 1958, the Texas Instruments laboratories on North Central Expressway in Dallas emptied out. The company observed a two-week summer break. The cafeterias closed. The senior engineers went to the lakes, the families packed station wagons, the parking lots thinned to almost nothing. Almost. A new hire named Jack Kilby, six feet six inches tall and in his second month on the job, had not accumulated any vacation days. So he came to work alone.

Kilby was thirty-four. He had grown up in Great Bend, Kansas, the son of a small-town power-company executive who worked weekends with a ham radio when blizzards knocked out telephone lines. He had failed the entrance exam to MIT, settled for the University of Illinois, served as an Army Signal Corps radio operator in Burma during the war, and spent the next ten years at a small Milwaukee component-maker called Centralab, where he had developed silk-screened ceramic substrates and earned his first patent. He was not the kind of engineer who turned heads in a meeting. At six feet six inches he was hard to miss in any room, but his colleagues remembered him for the opposite quality. He was slow-spoken, mild, almost shy, the sort of man who could stand in front of a blackboard for an hour and never raise his voice. When he later won the Nobel Prize, his colleagues at TI would describe him in the same terms friends had used in Kansas: a quiet Midwesterner who happened to think in three dimensions. Years afterward, the company’s Dallas campus would call him the humble giant, and mean it both ways.

He had arrived in Dallas on May 15, 1958, recruited by a TI manager named Willis Adcock to attack a problem the U.S. Department of Defense had been pouring money at for years and getting almost nowhere. Engineers called it the tyranny of numbers. The transistor, that small slab of doped germanium or silicon Bell Labs had unveiled a decade earlier, had begun to multiply. A naval shipboard radar in 1958 might require fifty thousand discrete electronic components, each one wired to its neighbors by hand, each solder joint a potential failure point. The Army’s pet program, called Micro-Module, tried to attack the problem by cramming standardized component wafers into stacks. Kilby thought it was a dead end. So did most of TI’s senior staff. But TI had bid on the work, and somebody needed to figure out what to do.

In the empty lab, Kilby began sketching. Through the second and third weeks of July he kept returning to the same insight: nobody had thought to ask whether the components themselves had to be different things. A resistor and a capacitor and a transistor were each, at heart, a configuration of doped semiconductor. If they were all made of the same material, why mount them on a circuit board with copper traces between them? Why not build the entire circuit, transistors and resistors and capacitors and the wires too, out of a single sliver of germanium? On July 24, in the kind of hand that engineers used in laboratory notebooks, he drew it. He titled the entry “Extreme miniaturization of electronic circuits.” He set down five components, all formed within one piece of semiconductor.

When Adcock came back from vacation, Kilby took the sketches in. Adcock listened. He was, by Kilby’s own later description, dubious. He told Kilby to prove it. Build something. Build the simplest version of the thing he had drawn, a phase-shift oscillator, and demonstrate it.

On the morning of September 12, 1958, Kilby took a sliver of germanium roughly seven sixteenths of an inch long and one sixteenth wide, with all of his components etched and diffused into it, and connected the leads with a few hair-thin gold wires. He carried it down the corridor to a lab where Adcock and Mark Shepherd, the head of TI’s semiconductor division, and a handful of other executives had gathered around an oscilloscope. Kilby plugged in the germanium chip. He turned on the power supply. A clean, continuous sine wave traced itself across the green phosphor of the oscilloscope screen.

There is no record of cheering. The men in the room had spent decades around oscilloscopes. They had watched a thousand sine waves trace themselves across a thousand screens. What they understood, slowly, was that the wave they were looking at was being generated by a single piece of germanium the size of a paper clip. Every transistor, every resistor, every capacitor was inside that one slab. There were no breadboards. There were no jumper wires. There was just the chip. The tyranny of numbers had a way out.

Three weeks later, Kilby demonstrated a second device, a flip-flop trigger built the same way. By February 6, 1959, TI’s lawyers had filed his patent application. The title was “Miniaturized electronic circuits.” The drawings were so crude, with their flying gold wires looping awkwardly above the surface, that years later a federal court would seize on them. Patrick Haggerty, the TI president whose insistence on shipping the first transistor radio in 1954 had given the company its commercial swagger, wanted Kilby’s invention announced as soon as possible. On March 6, 1959, at a press conference in New York timed to coincide with the Institute of Radio Engineers convention, TI unveiled what it called the solid circuit. The company called it the most significant development in electronics since the silicon transistor. Most of the trade press paid polite attention. Most of the buyers shrugged. The thing was expensive, fragile, and unproven.

In Mountain View, California, more than fifteen hundred miles from Dallas, a thirty-one-year-old physicist named Robert Norton Noyce read the announcement and recognized, in his quiet way, that he had been working toward the same thing.

Noyce was the third of four sons of a Congregational minister. He had grown up in Grinnell, Iowa, a college town set in cornfields, where his father preached and his mother kept the books for both house and parish. He was charming and athletic and almost compulsively confident. As a college junior he had stolen a piglet from a local farmer for a luau and nearly been expelled, talked his way out with the help of his physics professor Grant Gale, and then, when Gale obtained two of the world’s first transistors from Bell Labs in 1948, found his life’s subject. He went to MIT for his doctorate, took a job at Philco in Philadelphia, and in 1956 accepted an offer from William Shockley to come west and work on transistors in California.

He lasted at Shockley Semiconductor barely a year. Shockley was a Nobel Prize winner with a paranoid streak who recorded telephone calls and threatened lie-detector tests. He had pivoted the lab away from the silicon transistors his hires wanted to build and toward a four-layer device of his own design that nobody else thought practical. In May 1957 a small delegation took its grievances to the company’s parent, Beckman Instruments, hoping the founder Arnold Beckman would replace Shockley. Beckman declined. By June, seven of Shockley’s brightest hires had decided to leave on their own. Noyce, the youngest manager and the one Shockley most trusted, joined them late, only after the others had committed.

Shockley called them the Traitorous Eight, a label they would wear for the rest of their lives. Through Eugene Kleiner’s letter to a Hayden, Stone broker in New York they had picked up two financiers, Arthur Rock and Bud Coyle, and through Rock and Coyle they had picked up Sherman Fairchild, the IBM heir whose camera-and-instrument company agreed to lend them $1.38 million. On September 18, 1957, they walked out of Shockley Semiconductor and started Fairchild Semiconductor in a low concrete-block building at 844 East Charleston Road in Palo Alto, on a quiet stretch of orchards south of San Francisco. Noyce, at twenty-nine, became director of research and development.

What followed at Fairchild was the more important set of inventions, because they made manufacturing possible. The Eight brought commercial silicon transistors to market within a year. Gordon Moore, the youngest of the founders, ran the diffusion process. Sheldon Roberts grew the crystals. Hoerni worked on devices. Noyce did everything from running the company to walking samples down to the testing benches and back. By the early summer of 1958, IBM’s Federal Systems Division had ordered one hundred of the new transistors at $150 apiece, intended for the avionics of the B-70 supersonic bomber. No established maker could meet IBM’s voltage and reliability specifications. The Eight could. They named the device the 2N697, demonstrated it at the Wescon trade show in August, and unveiled it to the world the same month Kilby was sketching his single-slab oscillator in Dallas. By the time Kilby demonstrated his germanium chip on September 12, Fairchild was already shipping silicon devices and looking for its next contract.

The decisive Fairchild breakthrough came not from Noyce but from Jean Hoerni, a Swiss-born physicist with a doctorate from Cambridge who had been one of the eight. Hoerni had been complaining for over a year about the way semiconductor surfaces were being processed. The standard practice was to grow a layer of silicon dioxide on the wafer to mask the diffusion of impurities, then to etch most of it away to leave the bare silicon exposed. Everyone did it that way. Hoerni thought it was an act of vandalism. The oxide, he believed, was protecting the delicate junctions underneath from contamination. As early as December 1957 he had jotted into his notebook the heading, “Method of protecting exposed p-n junctions at the surface of transistors by oxide masking.” Noyce had countersigned the entry and then gone back to other work. The idea sat for over a year.

In January 1959, Hoerni finally built one. He left the oxide on, deposited aluminum contacts through small windows, and produced a transistor that was not only cleaner but also flatter than anything anyone in the industry had seen. The whole structure, instead of climbing up off the wafer in mesa-shaped lumps, was planar. He filed his patent disclosure that same month. By March he had a working device. By April 1960, Fairchild was selling it commercially as the 2N1613.

Hoerni’s planar process is the most consequential single invention in this whole story. The integrated circuits Kilby and Noyce had imagined as a single sliver of semiconductor needed a way to electrically isolate components from one another, and they needed a way to wire those components together without flying gold leads. The planar process gave them both. The leftover oxide layer was an electrical insulator. You could deposit a metal film right on top of it, photolithographically pattern it into thin lines, and use those lines to connect the diffused regions underneath, like a printed circuit board grown out of the silicon itself.

Noyce understood this almost immediately. On January 23, 1959, on pages 70 and 71 of a Fairchild laboratory notebook, he wrote out an idea he titled “Methods of isolating multiple devices in semiconductor wafers.” He described how Hoerni’s planar transistor could be extended: instead of cutting the wafer into individual transistors, leave many of them on a single piece of silicon, isolate them from one another with reverse-biased p-n junctions in the substrate, and connect them with patterned aluminum lines deposited over the protective oxide. The notebook page is not pretty. The drawings are workmanlike. The handwriting is steady. But what Noyce had described, in those two pages, was the architecture of nearly every chip the world has built since.

Fairchild filed the patent on Noyce’s “Semiconductor Device-and-Lead Structure” on July 30, 1959, more than five months after TI had filed Kilby’s. Jay Last, another of the Eight, ran a small team that fabricated the first working monolithic IC in the planar style by September 1960. In March 1961, at the IRE convention in New York, Fairchild went public with the result and gave it a brand: Micrologic.

The two inventions were close enough in time, and close enough in concept, that lawyers at TI and Fairchild spent the next decade fighting about which was first. The U.S. Patent Office declared an interference, an internal proceeding to determine priority. Both sides assembled witnesses, notebooks, and dueling expert affidavits. The Board of Patent Interferences split the difference, awarding some claims to Kilby and others to Noyce. Both companies appealed. On November 6, 1969, the U.S. Court of Customs and Patent Appeals issued its ruling. The judges focused on a single phrase from Kilby’s patent specification, which had described conductive material being “laid down” on the chip. Did the term “laid down” cover the kind of vapor-deposited, photolithographically patterned aluminum film Noyce had specified? TI argued yes. The court disagreed. The wording, it concluded, did not extend to Noyce’s metallization. The broader integrated-circuit claims went to Noyce. By that point neither side particularly cared. Three years earlier, in the summer of 1966, Fairchild and TI had quietly signed a cross-licensing agreement. They were already sharing royalties. The chip industry would treat both men as co-inventors for the rest of their lives.

The legal outcome mattered less than the divergence the two patents represented.

TI’s chip was the first. TI’s chip was also a hand-built, gold-wired curiosity, designed to demonstrate a principle rather than to be manufactured at scale. Its biggest customer was the U.S. military. In the fall of 1962, Texas Instruments won a contract from Autonetics to develop and supply twenty-two custom integrated circuits for the guidance computer of the Minuteman II intercontinental ballistic missile. By 1965, the Minuteman program had become the largest single buyer of integrated circuits in the world, surpassing even NASA. The Department of Defense was paying premium prices for parts whose chief virtue was reliability under acceleration, vibration, and radiation. TI’s strength under Pat Haggerty had always been the willingness to take government contracts, deliver under tight specifications, and let federal procurement underwrite the early years of the learning curve. The integrated circuit slotted into that culture without friction. The first generation of TI chips went into nuclear-tipped missiles aimed across the Arctic Circle.

Fairchild’s chip, by contrast, was designed from the first to be made in volume. Hoerni’s planar process meant that hundreds, eventually thousands, of identical devices could be diffused, oxidized, photolithographically patterned, and metallized on a single wafer at the same time. In March 1961, when Fairchild rolled out the Micrologic family, it priced its first commercial integrated circuit at $120 per chip. That was an absurd price compared with the cost of wiring up the same logic from discrete transistors. But Noyce understood something the rest of the industry was slow to grasp. The price would fall, fast, because the manufacturing process had no fundamental floor. Each generation of equipment would shrink the line widths, raise the yields, and let Fairchild build more chips per wafer. The early Micrologic gates went to NASA. The Apollo Guidance Computer, designed at MIT and built by Raytheon, used thousands of Fairchild three-input NOR gates per machine. By 1965, Fairchild had cut its IC prices to a few dollars apiece. By 1968, a Micrologic gate that had cost $120 in 1961 was selling for about a dollar.

Inside the two companies, the cultural difference was visible to anyone who walked the halls. Fairchild’s headquarters in Mountain View became a kind of free port. Noyce sat at a secondhand metal desk in an open bullpen, refused reserved parking spaces, dressed in shirtsleeves, and gave his managers their problems and stepped back. Stock options went down to the technicians. Decisions were made in the corridors, often by whoever cared most about the problem at hand. Engineers came in young, did the most important work of their lives, and a few years later left to start companies of their own. Within a decade, Fairchild’s alumni would seed dozens of firms across what was not yet called Silicon Valley. The model of how to run a chip company, and within a few years the model of how to run a technology company at all, came out of that one Palo Alto bullpen.

TI in Dallas was a different kind of place. It was a defense contractor, a survivor of the Texas oilfield-services boom, run by men in white shirts and dark ties who had won their stripes building geophysical instruments for Shell and Magnolia and acoustic mines for the Navy. The discipline was tighter, the chain of command longer, the relationship to government customers older and closer. Pat Haggerty, the company’s president, did believe in chasing commercial markets, and he had pushed TI into the Regency transistor radio in 1954 specifically to prove the point. But the institutional center of gravity in Dallas always tilted toward Washington. The big checks came from the Pentagon and from NASA. The product specifications came in MIL-STD documents. The contracts were sealed across conference tables in Arlington and Houston, not in venture meetings on Sand Hill Road.

Kilby thrived in that environment in his way. He stayed at TI for twelve more years, led the team that designed the first handheld electronic calculator in the late 1960s, and contributed to a long string of TI patents. He left full-time in 1970 to work as a private inventor, living in Dallas, attending the camera club, photographing his daughters, and rarely speaking publicly about the chip. He was famously laconic. At his 1982 induction into the National Inventors Hall of Fame, his entire acceptance speech consisted of two words: thank you. When he was awarded the Nobel Prize in Physics in 2000, his Stockholm lecture opened by attributing his success to the freedom TI had given him in the summer of 1958, and went out of its way, more than once, to credit Noyce for the work at Fairchild that had made the integrated circuit a manufacturing reality. By then Noyce had been dead a decade.

Noyce never won the Nobel. He had said, in his usual disarming way, that they didn’t give Nobel Prizes for engineering, and he had moved on. In 1968, eleven years after walking out of Shockley’s lab, he and Gordon Moore walked out of Fairchild too, taking a young chemical engineer named Andy Grove with them, to start a small company that would eventually be called Intel. By the time he died of a heart attack at sixty-two, in Austin in 1990, Robert Noyce had done more to shape the commercial chip industry than any other living person.

The fork in the road that began in 1958 and 1959 was never quite closed. The American chip industry would always have two parents, and the parents had different temperaments. From TI’s side came the model of the chip as strategic instrument, paid for by the Pentagon, designed for environments where reliability mattered more than cost, valued for what it could do inside missiles and submarines and bombers. From Fairchild’s side came the model of the chip as commercial product, sold by the millions, priced down a learning curve, built for whoever wanted to buy. The first model funded the early decades. The second model, in the long run, ate the world.

It was hard to see that distinction in 1958, watching Kilby’s sine wave wobble across an oscilloscope in an empty Dallas lab. It was almost as hard to see it in January 1959, watching Noyce’s pen move across two pages of a Fairchild notebook. The men involved would have been the first to say so. Hoerni would later complain, with some bitterness, that history was simplifying a story that was really five or six people’s. Kilby would shrug off most of the credit and say his contribution had been the smallest practical part of the larger problem. Noyce, charismatic to the end, would joke that he was just the marketer. None of them spoke as if they understood that what they had done was build the foundation of an industry that would, within four decades, sit at the center of every economy and every army on the planet.

They did not need to understand it. They had built the thing. The next twenty years would belong to the engineers and salesmen who figured out what to do with it.