In 1965, Gordon Moore was asked to contribute a prediction to the thirty-fifth anniversary issue of Electronics magazine. He was then director of research and development at Fairchild Semiconductor, and he was asked what the next ten years would hold for chip components. His short article, “Cramming more components onto integrated circuits”, noted that the number of components on a circuit had been doubling every year and expected that to continue for at least a decade. Looking back in a 2015 interview, he called it “a wild extrapolation”.
That extrapolation became the most quoted rule of thumb in computing. It was never a law of nature, only an empirical observation, the kind of experience curve in which production gets cheaper as makers learn. Yet it has held since 1975, and the name stuck.
From one year to two
At the 1975 IEEE International Electron Devices Meeting, Moore revised his forecast. Complexity would keep doubling annually until about 1980, then slow to a doubling roughly every two years, a compound annual growth rate of 41 percent. He credited three things: metal-oxide-semiconductor technology, larger dies combined with fewer defects, and what he called “circuit and device cleverness”.
Shortly afterwards Carver Mead, a Caltech professor, popularised the phrase “Moore’s law”. Chipmakers adopted it as a target, and some saw it as a self-fulfilling prophecy. Moore himself thought it surprising and optimistic: “Moore’s law is a violation of Murphy’s law. Everything gets better and better.”
The figure most people remember, 18 months, is a misquote. It comes from David House, an Intel executive, who reasoned in 1975 that twice the transistors every two years meant chip performance doubling about every 18 months, with no extra power. That second part leaned on Dennard scaling, formulated by Robert Dennard of IBM in 1974, which held that smaller transistors kept their power density constant.
What the rule did for ordinary lives
The industry used the prediction to plan long-term and to set research targets. Cheaper microprocessors, bigger memory, better sensors and higher pixel counts in digital cameras are all strongly linked to it, and together they have driven technological and social change, productivity and economic growth. For families across the region, that is the story of the phone in the pocket getting better every year without costing more.
The slowdown
The squeeze came in two stages. Dennard scaling broke down in the mid-2000s. Then, since around 2010, microprocessor architects say, semiconductor advancement has run industry-wide at a pace slightly below Moore’s prediction. Brian Krzanich, Intel’s former chief executive, pointed to the 1975 revision as a precedent, calling the deceleration “a natural part of the history of Moore’s law”. Much of the industry has since turned to the needs of major computing applications instead of pure scaling.
Costs rise as the transistors shrink. The tools used to make chips, principally extreme ultraviolet lithography, double in cost every four years. That fed what is known as Moore’s second law, or Rock’s law, after Arthur Rock: the capital cost of a fabrication plant also grows exponentially.
Still, as of 2019, TSMC and Samsung Electronics claimed to be keeping pace, with 10, 7 and 5 nanometre nodes in mass production. Experts have not agreed on when the law will end. In September 2022, Nvidia chief executive Jensen Huang considered it dead, while Intel’s then-chief Pat Gelsinger held the opposite view.
