Gordon Moore Has Passed Away At The Age Of 94

Gordon Moore died on March 24, 2023, at the age of 94. His name is inseparable from one of the most consequential observations in the history of technology: Moore’s Law. The observation shaped not only the semiconductor industry but the software written to run on it, and the expectations of everyone who used computers for half a century.

The 1965 paper

In April 1965, Moore — then director of research and development at Fairchild Semiconductor — published a short article in Electronics Magazine titled “Cramming More Components onto Integrated Circuits.” The article was not a scientific paper. It was an industry forecast.

Moore observed that the number of components that could be placed on an integrated circuit had roughly doubled each year since integrated circuits were invented, and that this trend would likely continue “for at least ten years.” In 1965, the most complex chip available had about 50 transistors. Moore was predicting a path toward 65,000 components by 1975.

He was right about the direction, though the exact doubling period was revised. By 1975, Moore updated his observation: doubling every two years was more accurate than annually. David House, a colleague at Intel, later popularized the version most people know — doubling every 18 months — which combined the transistor count and the speed improvement.

The self-fulfilling mechanism

What makes Moore’s Law unusual as a prediction is that it influenced the thing it described. The semiconductor industry organized its research roadmap around the expectation of continued transistor scaling. The International Technology Roadmap for Semiconductors, updated every two years by major industry players, set targets based on Moore’s projection. Suppliers, toolmakers, and manufacturers all planned their investments around the expectation that if they did not hit the next node by a certain date, they would fall behind.

This created a coordination mechanism. Because everyone expected the doubling to continue, everyone invested to make it continue. A prediction that becomes a target and then a production plan operates differently from a prediction made by an outside observer.

For software, the effect was profound. Programmers wrote code with the knowledge that hardware would be faster in two years. Problems too expensive to solve today could be deferred to the next chip generation. Entire approaches — video compression, real-time graphics, database engines, operating systems — were designed around the assumption of predictable hardware improvement.

Where the law runs into physics

As of 2023, transistor scaling has not stopped, but it has become enormously more expensive and technically difficult. Modern chips from TSMC and Samsung use process nodes measured in nanometers — 3nm and 2nm are in production or development — which describes features smaller than some proteins.

At these scales, quantum effects become significant. Electrons can tunnel through barriers that are only a few atoms thick. Managing heat becomes more complex. The economic cost of each process generation has increased sharply: a new fabrication plant now costs $20 billion or more.

The industry has responded with approaches beyond simple transistor scaling: 3D chip stacking (multiple layers of silicon bonded together), chiplet architectures (assembling specialized dies into a single package, as AMD does with its Ryzen processors), extreme ultraviolet lithography using light at 13.5nm wavelength to pattern ever-smaller features. These methods continue to improve computing capability, but through different mechanisms than the transistor scaling Moore originally observed.

What remains

Gordon Moore was also a co-founder of Intel with Robert Noyce in 1968. Intel became the dominant manufacturer of microprocessors for personal computers, and Moore’s leadership shaped its early direction. His later philanthropy — through the Gordon and Betty Moore Foundation, established in 2000 — focused on science, environmental conservation, and healthcare.

What stays with me is the reminder that the expectations embedded in a technology have consequences as real as the technology itself. For decades, software could count on faster hardware arriving on schedule. That shaped what was built and how. As the pace of transistor scaling slows, the industry is finding new paths forward. But the culture Moore’s prediction created — the expectation that computing will improve, the confidence that tomorrow’s hardware will expand what today’s software can do — is his more durable legacy.