A Casual Guess That Became Gospel
In 1965, an engineer named Gordon Moore sat down to write an article for a trade magazine. He noticed something odd: the number of components squeezed onto an integrated circuit had been doubling every single year. So he made a guess. He said this trend would probably keep going for at least another decade.
Moore later admitted it wasn't rigorous science. "I just did a wild extrapolation," he said in 2015, looking back on his own famous prediction. Ten years later, in 1975, he revised his own guess, saying the doubling would happen roughly every two years instead of every year, a pace known as a 41% compound annual growth rate.
A Caltech professor named Carver Mead gave the idea its lasting name shortly after: Moore's law. Despite the label, it was never a law of physics. It was one engineer's hunch about how fast his industry could move.
The Industry Built Its Roadmap Around One Man's Napkin Math
Here's the strange part: Moore's law didn't just describe the semiconductor industry. It started running it. Chipmakers began treating the prediction like a deadline. Engineers set their research goals around hitting the next doubling, because falling behind meant falling behind everyone else too.
An industry group called the International Technology Roadmap for Semiconductors used Moore's law as its central organizing principle starting in 1998, coordinating research across companies worldwide. It only abandoned that approach in 2016, switching to a strategy based on what applications actually needed instead.
Even the popular "18 months" version of Moore's law isn't quite what Moore said. An Intel executive named David House calculated that number in 1975, combining smaller transistors with faster switching speeds. Moore's own number was always two years.
The Price Tag Nobody Talks About
While consumers enjoyed cheaper and cheaper chips, someone else was footing a very different bill. Building the factories that make those chips got dramatically more expensive over time, even as the chips themselves got cheaper per transistor.
This pattern earned its own nickname: Moore's second law, also called Rock's law after investor Arthur Rock. It holds that the capital cost of a semiconductor fabrication plant roughly doubles every four years. Today, a single cutting-edge chip factory can cost billions of dollars to build.
That rising cost is quietly one of the biggest threats to Moore's original prediction. It's not just about whether engineers can keep shrinking transistors. It's about whether any company can afford the extreme ultraviolet lithography equipment needed to build the smallest chips at all.
The Slowdown Nobody Wanted to Admit
For decades, Moore's law felt unstoppable. Then, around 2010, chip architects across the industry started noticing something uncomfortable: progress had quietly slowed below the pace Moore predicted. A key assumption called Dennard scaling, which said power use would stay constant as transistors shrank, had already broken down in the mid-2000s, making heat a serious problem.
The numbers tell the story. Single-core processor performance improved by about 52% a year between 1986 and 2003. Between 2003 and 2011, that fell to 23% a year. By 2018, it had dropped to just 7% a year.
Even Intel admitted it. In 2015, then-CEO Brian Krzanich said the company's pace had become "closer to two and a half years than two." By 2023, Intel's Pat Gelsinger put the real number closer to three years.
Is Moore's Law Already Dead?
Ask two chip executives whether Moore's law is still alive, and you'll get two different answers. In September 2022, Nvidia CEO Jensen Huang flatly declared Moore's law dead, using it to justify higher prices on gaming graphics cards. Intel's then-CEO Pat Gelsinger disagreed, insisting the law still had life left in it.
Gordon Moore himself saw this coming decades earlier. Back in 2005, he said plainly: "It can't continue forever. The nature of exponentials is that you push them out and eventually disaster happens." He guessed the industry had 10 to 20 years before hitting a fundamental wall, pointing to the size of atoms themselves as the ultimate limit.
Transistors have already shrunk to almost unimaginably small sizes, moving from a 7 nanometer process in 2018 to 3 nanometers by 2022. There isn't much room left to shrink further.
