Chapter 3 · 1956–1971

West, and the promise

A bad boss, eight defectors, and a magazine article that turned out to be a law of nature.

The same year the decree opened the patents, William Shockley went home. He was the transistor’s most brilliant and most difficult co-inventor, and he’d concluded Bell Labs would never make him rich or put him in charge. So in 1956 he moved back to Palo Alto, where his mother lived, and started Shockley Semiconductor. Because he was famous, he could recruit the best young scientists in the country. Because of who he was, he couldn’t keep them.

He managed by suspicion. He posted salaries. He ordered lie-detector tests. Within a year, eight of his best — Robert Noyce and Gordon Moore among them — quit together. Shockley called them the traitorous eight, and the name stuck because it was half true: they didn’t just leave, they took the mission. With backing arranged through a young financier named Arthur Rock, they founded Fairchild Semiconductor in 1957.

Understand what that walkout invented, beyond chips. Leaving the boss to start the rival wasn’t done in 1957. Fairchild made it the founding ritual of an entire region — over the next decades, Fairchild alumni started dozens of companies, and the money model behind the walkout became venture capital. The decree put the transistor in anyone’s hands; the defection built the culture that did something with it. Geography is destiny’s punchline here: the whole thing happened where it did because Shockley’s mother lived in Palo Alto.

At Fairchild, Noyce co-invented the integrated circuit — many transistors made together on one piece of silicon, wiring included. No more soldering individual transistors like beads. You printed circuits the way you print a page. And once you’re printing, the question becomes: how small can the print get?

Gordon Moore answered it in 1965, in a trade-magazine article he was asked to write about the future of electronics. He counted the components on each year’s best chips and noticed the count doubled roughly every year — later revised to every two. Then he did the audacious thing: he drew the line forward and predicted it would keep doubling. It read like an observation. It behaved like a promise. The industry started planning around the doubling, which made the doubling come true, which made the planning safer. For fifty years, Moore’s law was less physics than a schedule the entire industry agreed to keep.

Feel what a doubling schedule means before you trust any summary of it. The slider below runs from 1971 to now. It starts at 2,300 transistors — a small high school’s worth of students — and ends north of a hundred billion, the stars of the Milky Way. Same size chip. That is not improvement. That is a different universe arriving on a payment plan.

In 1968, Noyce and Moore defected a second time — from their own company — and founded Intel, bringing along a ferociously intense Hungarian-born engineer named Andy Grove, who had escaped Budapest on foot at twenty and ran operations like it could all be taken away tomorrow. Intel made memory chips. Then in 1969 a Japanese calculator company asked for a dozen custom chips, and an Intel engineer named Federico Faggin — newly arrived from Italy — compressed the whole job into one: a complete computer brain on a single sliver of silicon. The Intel 4004, 1971. The first commercial microprocessor. It was sold to run a desk calculator.

Look at the roster of this chapter. Grove from Hungary. Faggin from Italy. Later chapters add Finland, Taiwan, Ukraine, France. The valley the decree built was staffed from everywhere — that thread runs the length of this book without needing a speech.

So by 1971 the promise is in place: silicon that doubles on schedule, an industry organized around the doubling, and a computer that fits on a chip. The silicon branch of the map is live. The next two chapters cross back to the software branch — because while Fairchild was printing circuits, two men at Bell Labs were about to lose their giant computing project and build something small instead.