In 1965, a researcher at Fairchild Semiconductor had a chip with 64 transistors on it in his lab. A year earlier the best chip held 32. Gordon Moore plotted the points, saw a pattern, and wrote down a guess that would steer an entire industry.
The guess appeared in the 19 April 1965 issue of Electronics magazine, under the title “Cramming more components onto integrated circuits”. Moore had first written it as an internal document to persuade his company’s customers to build with the newest chips. Drawing a line through just five data points, he projected that by 1975 economics might dictate squeezing as many as 65,000 components onto a single silicon chip. That amounted to a doubling every 12 months. Venture capitalist Steve Jurvetson later called the graph in that article “the most important graph in human history”.
An observation, not a law
Moore never saw it as a law of nature. ExtremeTech notes that he was the first to plot the points and make a specific forecast, though he got the idea of rapid miniaturisation from hearing Douglas Engelbart speak. His reasoning rested on cost. More components on a chip meant a lower cost per component, but yield fell as the count rose, so there was an optimal number at any moment, and that number was growing fast.
The name came later. Carver Mead, a Caltech professor and Moore’s friend, noticed that the trend had held for the decade and gave it the name Moore’s Law.
The rate changes
By 1975 Moore, then a co-founder at Intel in charge of R&D, told an IEEE meeting that better photolithography, larger wafers, improved process technology and “circuit and device cleverness” had made his projection come true. He then lowered his estimate to “a doubling every two years, rather than every year”. Intel’s Dave House added another twist, arguing that because components were also getting faster, computing power on a chip could double about every 18 months. That version became the one the semiconductor industry tracked most closely.
In 1995, when an Intel Pentium held nearly 5 million transistors, Moore concluded: “The current prediction is that this is not going to stop soon.” The Computer History Museum records that AMD and Intel chips stayed on the curve through 2010, when products from both companies held more than one billion transistors.
What it meant, and the question of an end
The museum quotes David Brock of the Chemical Heritage Foundation: the number of transistors made in a single year matches or exceeds the grains of sand on all the world’s beaches. The price of computing has fallen over a million-fold, and the cost of electronics a billion-fold. For ordinary buyers, that is the reason phones and laptops keep getting more capable without costing more.
Moore himself warned in 2003 that “No exponential change continues forever.” When the museum marked the 50th anniversary in 2015, a survey found that only one fourth of semiconductor business leaders believed the law would last for the foreseeable future. Intel’s Mark Bohr, though, described plans to move from chips with 14 nanometre features to 7 nanometres, which suggested the law would hold at least to the end of that decade. ExtremeTech’s view was that the integrated circuit was running into limits of size and power, with parallel computing on GPUs and, further out, quantum computing waiting as possible successors.
Sources
- Moore’s Law@50: “The most important graph in human history” (Computer History Museum)
- Moore's Law at 50: Its past and its future (ExtremeTech)
