How did one company become two thousand?

The compensation structure was the problem

The deal Sherman Fairchild had given the eight founders was a buyout option: after a few years, Fairchild Camera and Instrument could acquire the subsidiary outright for $3 million, at which point the founders would split the proceeds. In 1958 that looked like a lot of money. By 1965, Fairchild Semiconductor was earning more revenue than Fairchild Camera, and the option had been exercised — the founders had been paid out and were now employees of a company they had built, with no further equity upside.

Beneath them, the next layer of senior engineers had no equity at all. They were salaried employees of a New York-controlled subsidiary, watching their work translate into balance-sheet entries that benefited shareholders they had never met. The corporate parent extracted profits from Mountain View to fund unrelated businesses on the East Coast, including the floundering aerial-camera arm the company was actually named for. Engineers who had invented the silicon planar transistor and the integrated circuit were paid the same as engineers at Westinghouse.

This was not an accident. Sherman Fairchild's entire investment thesis was to harvest a profitable subsidiary, not to share upside with operators. The structure was conventional for a 1950s East Coast conglomerate. What was unconventional was the geography it had been planted in — and the kind of person it had recruited. The engineers who had been willing to quit Shockley were not the kind of engineers who would sit still for a corporate cash extraction strategy. They left.

The first defections

The leaving started slowly. In 1961, Jay Last, Jean Hoerni, and Sheldon Roberts — three of the original eight — left Fairchild to form Amelco, a subsidiary of Teledyne. The exit was framed as a strategic disagreement about funding for new products, but the deeper problem was equity. Within a few years, several more senior people had left to start related ventures: Signetics, General Microelectronics, Molectro, Rheem Semiconductor.

Each of these spinoffs had the same shape. A senior Fairchild engineer or manager would identify a product line Fairchild was not prioritizing, find an investor willing to fund a new company around it, and recruit a few colleagues to come along. The new company would compete directly with Fairchild on adjacent products, often using process technology the founders had developed at Fairchild. Fairchild rarely sued. The non-compete enforcement environment in California made litigation expensive and unwinnable; the company simply absorbed the loss and kept training the next cohort.

By 1967, a Forbes article counted thirty Fairchild spinoffs. By 1970, the count was past fifty. The label "Fairchildren" started showing up in trade press the same year. It was sometimes used affectionately, sometimes ironically — but always in the same sense: Fairchild had become an inadvertent founder factory, and everybody knew it.

1968: the year the dam broke

In the summer of 1968, the two senior remaining founders — Robert Noyce and Gordon Moore — quit Fairchild together. Their stated reason was that the East Coast parent had blocked Noyce from being appointed CEO of Fairchild Semiconductor and had instead installed an outsider. The real reason was the same as everyone else's: they were running a billion-dollar business for the upside of strangers.

They walked out, called Arthur Rock (the same investor who had funded the original Fairchild raise eleven years earlier), and within three weeks had $2.5 million in committed capital for a new company called NM Electronics — quickly renamed Intel. Andy Grove, a Hungarian refugee who had become Fairchild's assistant director of R&D, joined them as the third employee. He was 32 years old.

Fairchild's response was to lose its CEO and several senior managers within a year. Jerry Sanders, the head of marketing at Fairchild Semiconductor, was fired by the new East Coast appointee. Sanders left in 1969 with seven colleagues and started Advanced Micro Devices — AMD — funded again partly through Rock-adjacent networks. Charlie Sporck, the operations chief, had already left in 1967 to take over National Semiconductor, which became a third major IC company built on Fairchild alumni and Fairchild manufacturing know-how. By 1971, Fairchild Semiconductor had lost most of its top operators to spinoffs that were now its direct competitors.

The receiving company on the other side — the one that bought what was left of the corpse — was Schlumberger in 1979, then National Semiconductor in 1987. By that point Fairchild was a brand name with declining market share. The accidental venture firm it had been operating for fifteen years had produced more enduring value than the parent ever recovered.

Source: Christophe Lécuyer, *Making Silicon Valley* (MIT Press, 2006); Tom Wolfe, "The Tinkerings of Robert Noyce" (Esquire, 1983); SEC filings of Fairchild Camera and Instrument, 1960-1979.

Why the East Coast did not match this

In the same fifteen years that Fairchild seeded sixty-five spinoffs, equivalent East Coast semiconductor employers — RCA's solid-state division, Sylvania, Raytheon, Westinghouse, Philco-Ford — seeded almost none. RCA in particular had been an early transistor licensee, employed thousands of semiconductor engineers, and had a sophisticated research lab in Princeton. By any technical measure RCA should have produced its own Intel and its own AMD. It produced neither.

The reasons are now well-documented and uncomfortable. First, Massachusetts and New Jersey and Pennsylvania all enforced non-compete clauses aggressively. An engineer who left RCA Princeton to start a competing semiconductor company faced a credible two-year injunction from working in the field. The same engineer in Mountain View did not. Second, East Coast employers paid better in absolute dollars but offered no equity. The trade was lifetime employment in exchange for capped upside — a deal that made sense in the postwar industrial economy and stopped making sense as the new industry produced its first generation of self-made paper millionaires. Third, the local venture-capital ecosystem on the East Coast was institutional and conservative; the West Coast version was personal and willing to bet on individuals who had quit their last job a week earlier.

The compounding effect over fifteen years was structural rather than incremental. Each Fairchild spinoff hired more engineers, which made the regional labor pool deeper, which made the next spinoff easier to staff, which made California more attractive to graduates leaving Stanford and Berkeley, which gave Fairchild and its peers more talent to hire, which produced more spinoffs. The same generation of engineers in Princeton or Waltham circulated within the same employers for decades. By 1980 the productivity differential was so wide that East Coast semiconductor employers began shutting their fab operations and outsourcing to California, then to Japan, then to Taiwan — every step away from the original site that the same conditions had set in motion.

Strategic read

A spinoff factory is not a happy accident. It is the outcome of four conditions stacked together: a generation of senior employees with no equity upside in their current employer; a labor-law environment that lets them leave without litigation; a local venture-capital network willing to fund them within days, not quarters; and a critical mass of similar companies dense enough that talent moves can happen across the street rather than across the country. Remove any one of the four and the rate of company formation drops by an order of magnitude. The other conditions — clever ideas, technical capability, capital availability in the abstract — matter much less than people assume.

The same pattern now describes frontier AI labs. OpenAI, Anthropic, Google DeepMind, and to a lesser extent Meta AI and xAI are running the Fairchild-equivalent labor pump. Senior researchers join, contribute to a generational product, recognize that the upside is concentrated in a small number of founders or early employees, and leave to start new companies that often compete directly with their former employer. The same four conditions apply: equity asymmetry between founders and the next layer down, California labor law, a venture capital network that responds within a fortnight, and density. The 2026 list of AI spinoffs from frontier labs already runs into the high dozens. By 2030 it will be in the hundreds, the way the Fairchildren were by the late 1970s.

The investment-grade question is which of those spinoffs will produce the next Intel, in the sense of becoming the dominant player in a new sub-industry. Most will not. A small number will. The historical pattern suggests that the spinoffs that win are usually not the ones doing the obviously parallel thing to the parent (an AI lab making another foundation model) but the ones doing the perpendicular thing the parent could not justify (an inference-stack company; a domain-specific application; a tools company; a vertically-integrated AI-native product). Intel was perpendicular to Fairchild — Fairchild was making transistors and ICs for everyone, Intel decided to make memory chips specifically for the new minicomputer market — and it was the perpendicularity that gave Intel room to grow before Fairchild noticed.