In 1965 the trade magazine Electronics asked Gordon Moore, then director of research and development at Fairchild Semiconductor, to predict the next ten years of semiconductor components. Working from a handful of data points, the chemist noticed that the number of components on a chip at the lowest cost per component was doubling every year. He extended the line and estimated that by 1975 engineers would fit 65,000 components on a single piece of silicon.

This episode explains why that guess held, how it was revised, and how it turned into a schedule the whole industry felt obliged to meet. In 1975 Moore slowed the pace to a doubling every two years, and Caltech professor Carver Mead gave the idea its name. The story then turns to the costs: fabrication plants priced at upwards of $20 billion, the end of Dennard scaling around 2005, electrons tunneling through barriers a few nanometers thick, and the search for what comes after silicon.

• The 1982 Osborne Executive opens the episode. A smartphone from 2007 weighed about 100 times less, cost about a tenth as much after inflation, and ran roughly 100 times faster.
• The popular 18-month figure came from Intel executive David House, who combined Moore’s two-year doubling with Robert Dennard’s finding that smaller transistors run faster.
• Rock’s law, named for venture capitalist Arthur Rock, holds that the cost of a chip fabrication plant rises exponentially over time.
• Eroom’s law, Moore spelled backward, describes drug development, where the cost of a new medicine roughly doubles every nine years.
• In 2022 Nvidia chief executive Jensen Huang declared Moore’s law dead, while chipmakers turned to multi-core processors, 3D stacking, and FinFET transistors.