Science history: Computer scientist lays out 'Moore's law,' guiding chip design for a half century — Dec. 2, 1964

At a low-key talk for a local professional society in 1964, computer scientist and chemist Gordon Moore laid out a prediction that would define the world of technology for more than 50 years.

In the presentation for The Electrochemical Society, titled, "The Evolving Technology of the Semiconductor Integrated Circuits," Moore predicted that the number of transistors on an integrated circuit would double every year.

The final version of this prediction would become known as "Moore's law," and it would drive progress in the semiconductor industry for decades.

Although it's called a law, it was a prediction based more on economic dictates and industry trends than on the physical laws of nature.

Moore was a director of research and development at Fairchild Semiconductors when he gave the talk, and his goal was ultimately to sell more chips. At the time, computers were gigantic machines that took up a whole room, and integrated circuits, known as microchips, had somewhat limited practical applications.

The silicon transistor, the workhorse that does calculations in a computer, had been invented just a decade earlier, and the integrated circuit, which allowed computers to be miniaturized, had been patented just five years earlier. In 1961, the electronics company RCA had built a 16-transistor chip, and by 1964, General Microelectronics had built a 120-transistor chip.

Moore witnessed this dramatic progress and noticed that a mathematical rule seemed to be governing that progress. This mathematical correlation was later given the name "Moore's law" by other people.

Although Moore laid out the principle to The Electrochemical Society in 1964, it gained widespread traction in April of the following year, when he was asked to write an editorial in Electronics magazine. In it, he boldly predicted that as many as 65,000 components could be squeezed onto a single chip — an unheard-of number at the time. It's a charmingly small-potatoes number now, given that in 2024, a company unveiled a 4 trillion-transistor chip.

In 1968, Moore would co-found the chipmaker Intel, where his doubling law would go from a casual observation to a motivation for innovation.

Despite its name, Moore's law was never an ironclad rule. In 1975, Moore downgraded the pace of progress to transistor doubling every two years, rather than every year. That more modest doubling rate would become the official Moore's law, which would hold for years after. This relentless drive toward more computing power and miniaturization is what enables virtually all modern electronics, from the personal computer to the smartphone.

For years, people predicted that the law would become outdated, but it proved remarkably resilient for quite some time.

"The fact that we've been able to continue [Moore's law] this long has surprised me more than anything," Moore said in an interview with The Electrochemical Society in 2016. "There always seems to be an impenetrable barrier down the road, but as we get closer to it, people come up with solutions."

However, eventually, the principle would no longer hold. It's not clear exactly when Moore's law became defunct. In its canonical form, the standard likely died in 2016, as it took Intel five years to go from the 14-nanometer-size technology to 10 nanometers. Moore saw this happen, as this was years before he died at the ripe old age of 94 in 2023.

Eventually, Moore's "law" had to peter out because it runs up against the actual laws of physics. As transistors became ever smaller, quantum mechanics, the physics that governs the very small, began to play an outsize role. The world's smallest transistors can face problems with "quantum tunneling," wherein electrons in one tiny transistor can tunnel into another, thereby allowing current to flow in transistors that should be in the "off" position.

As a result, chipmakers are looking at designing chips with new materials and new architecture. The next Moore's law may apply to quantum computers, which harness quantum mechanics as a feature, not a bug, of calculations.