On this page
Phones, laptops and the servers behind every AI tool depend on chips that hold billions of transistors. A single prediction made in 1965 shaped how fast that happened. This post explains what Gordon Moore actually wrote, how he revised it, the doubling arithmetic behind it and what his way of working can teach a student.
Who was Gordon Moore?
Intel's obituary for Gordon Moore gives the outline. He was born in San Francisco on 3 January 1929 and earned a PhD in chemistry from the California Institute of Technology in 1954. In 1956 he joined Shockley Semiconductor. In 1957 he co-founded Fairchild Semiconductor with Robert Noyce and six other colleagues, and in July 1968 Moore and Noyce founded Intel. He died on 24 March 2023, aged 94.
His training was in chemistry. At Fairchild, Intel notes, he and Noyce played central roles in the first commercial production of diffused silicon transistors and of integrated circuits.
What did Moore predict in 1965?
The Computer History Museum explains that Moore, then director of R&D at Fairchild, drew a line through five data points for chips made between 1959 and 1964. An edited version appeared in Electronics on 19 April 1965 as "Cramming more components onto integrated circuits". A reprint is in the Proceedings of the IEEE.
The key sentence in the article is: "The complexity for minimum component costs has increased at a rate of roughly a factor of two per year." He added that there was no reason the rate would not hold "for at least ten years", which "means by 1975, the number of components per integrated circuit for minimum cost will be 65 000."
Note the words "minimum cost". Moore was talking about economics: the number of components at which each one is cheapest to make. The same article foresaw "home computers—or at least terminals connected to a central computer".
Why did he change it in 1975?
At the 1975 IEEE International Electron Devices Meeting, the museum says, Moore found his projection had been met. Looking at newer data that included microprocessors, he slowed the forecast to "a doubling every two years, rather than every year." The name "Moore's law" was coined by Carver Mead. By 1995, the museum notes, an Intel Pentium processor held nearly 5 million transistors.
Worked example: doubling every year or every two years
Let N₀ be the starting count and T the doubling time. After t years, N = N₀ × 2^(t/T).
Question 1. Moore's 1965 forecast doubled every year for 10 years. By what factor does the count grow?
2^(10/1) = 2^10 = 1,024. Working backwards, 65,000 ÷ 1,024 ≈ 63.5, so the forecast implies roughly 64 components in 1965. Check: 64 × 1,024 = 65,536.
Question 2. With the 1975 version, doubling every two years, what is the factor over the same 10 years?
2^(10/2) = 2^5 = 32. Same decade, a growth factor 32 times smaller than 1,024. That is why the revision mattered.
Question 3. How long does a factor of 1,000 take at one doubling every two years?
2^10 = 1,024 ≈ 1,000, so you need about 10 doublings, which is 10 × 2 = 20 years.
What mistakes do students make with doubling?
Adding instead of multiplying. Starting at 64 and adding 64 a year for 10 years gives 704, not 65,536. Growth by a fixed factor is exponential.
Mixing up the doubling time and the rate. Doubling every two years is not "100% a year"; it is about 41% a year, since √2 ≈ 1.41.
Changing what is being counted. The 1965 forecast was about the number of components on a chip, not about processor speed. Check the quantity before you quote the law.
Why does a chip forecast matter to physics students?
Every component Moore counted is a semiconductor device. If you want to see what a transistor does as a switch and an amplifier, start with our post on transistor basics. For the invention that made such chips possible, read about Jack Kilby and the integrated circuit.
Moore described his goal plainly in a 2008 interview quoted by Intel: "All I was trying to do was get that message across, that by putting more and more stuff on a chip we were going to make all electronics cheaper."
What can a student take from how Moore worked?
Plot the data before you predict. His forecast came from five points drawn on a graph.
Say what you are measuring. He tracked components at minimum cost, not just the largest chip possible.
Make a forecast that can be checked. "65,000 by 1975" could be shown wrong. It was not.
Revise in public. In 1975 he changed the rate when new data called for it.
Exponents and logarithms carry the whole argument. If they feel shaky, the Fermi Concept Library covers every concept in physics, chemistry, biology, mathematics and computing, so you can rebuild the idea before tackling problems.
Try it now with sample numbers: starting from 2,300 transistors, how many would you expect after 20 years of doubling every two years? (Answer: 2,300 × 2^10 = 2,355,200.)
