Moore’s Law Graph, 128 Years of Computing Power per Dollar

Key points from the graph

• The graph plots calculations per second per constant (inflation adjusted) dollar from 1900 to the mid 2020s, on a logarithmic scale where each tick is 100 times the one below.
• The vertical axis runs from 1E-09 (one billionth of a calculation per second per dollar) to 1E+13. Steve Jurvetson, who updates the chart, describes the span as a 1,000,000,000,000,000,000,000 times improvement in computation per dollar [1].
• The trend runs through five technologies in turn: mechanical, relay, vacuum tube, transistor and integrated circuit.
• Since about 2008 the highest points are no longer general purpose computers but NVIDIA GPUs, then custom chips (ASICs) such as Google’s TPUs and Tesla’s Dojo, and finally Mythic’s analog chips.
• The source line reads “Source: Ray Kurzweil, updated since 2008 by Steve Jurvetson, Future Ventures”.

The Moore’s law graph in text

The chart is titled “128 Years of Moore’s Law” and carries the Future Ventures astronaut logo. Each dot is one computer or chip. The horizontal axis is the year, marked every five years from 1900 to 2025. The vertical axis, “Calculations per second per constant dollar”, is marked 1E-09, 1E-07, 1E-05, 0.001, 0.1, 10, 1,000, 1E+05, 1E+07, 1E+09, 1E+11 and 1E+13. Shaded bands mark the five eras, and a legend gives the colours used inside the integrated circuit era. The chart prints no exact values for individual dots, so the table lists the labels in each era rather than readings.

There are also many unlabelled dots, mostly between 1945 and 1970 and between 1975 and 2010. Read from left to right, the dots form a rising band with no break where one technology replaces another. The band is flatter in the mechanical era and steeper after about 1950.

What the graph measures, and how it differs from Moore’s original law

Gordon Moore’s own prediction was about chips, not computers in general. In 1965, as Fairchild Semiconductor’s director of research and development, he plotted the number of components per integrated circuit and projected a doubling every 12 months. His article “Cramming more components onto integrated circuits” was published in Electronics on 19 April 1965. In 1975, by then at Intel, he slowed the rate to “a doubling every two years, rather than every year” [2]. Intel describes Moore’s law today as the observation that the number of transistors on an integrated circuit doubles every two years with minimal rise in cost [3].

This graph uses a different measure. Jurvetson explains that customers do not buy transistors but computational speed, so the chart tracks calculations per second for each constant dollar. That measure can be traced back long before the integrated circuit, which is why the chart starts with mechanical calculators. He says he first saw the idea in Ray Kurzweil’s 1999 book The Age of Spiritual Machines [1].

Jurvetson also notes that the popular figure of a doubling every 18 months is something Moore never said. In his words it is “a nice blending of his two predictions” from 1965 and 1975 [1].

Why the newest points are GPUs and custom chips

In his notes on the 2024 update, Jurvetson writes that leadership on the chart passed from Intel to NVIDIA about 15 years earlier, because the parallel design of a graphics processor suits the needs of deep learning. Custom chips built for neural networks take that further, and analog in memory computing, the approach behind the Mythic dots, goes further again. He adds that thousands of computers sit below the line, including everything from Intel in the past 15 years, because the chart shows only the frontier [1].

Notes on this infographic

• Not transistor counts. Despite the title, the chart does not plot transistors per chip, which is what Moore described [2]. It plots computing price performance, a broader measure.
• The top dot is a forecast. Jurvetson states that the top right dot for Mythic is “a prediction for 2026”, showing the effect of moving an older 40nm design to a newer process. It is not a measured product [1].
• No printed values. Individual dots carry names but no numbers, so positions can only be read approximately from the logarithmic scale.
• Spelling. The xAI Colossus label is printed “xAi Colossus”.
• Versions. Jurvetson has published several editions of this chart. This page shows the 128 year edition he posted on 3 December 2024 [1].

Questions about Moore’s law

What is Moore’s law?

It is the observation that the number of transistors on a chip doubles at a steady rate. Gordon Moore first predicted a doubling every year in 1965 and revised it to every two years in 1975 [2]. Intel states it as a doubling every two years with minimal rise in cost [3].

Is Moore’s law still valid?

Intel says it has a pipeline of research that gives it confidence it will maintain Moore’s law “for the next decade or longer”, and it is working towards 1 trillion transistors on a package [3]. Jurvetson, measuring computation per dollar rather than transistors, writes that the best planning assumption is that the trend will continue for the next 20 years as it has for the past 128 [1].

Is Moore’s law exponential?

Yes. A steady doubling over a fixed period is exponential growth. That is why the graph uses a semi-log scale, on which, as Jurvetson notes, a straight line is an exponential [1].

Sources

1. Steve Jurvetson, Future Ventures, The Moore’s Law Update, for 128 years, Flickr, 3 December 2024.
2. Computer History Museum, 1965, Moore’s Law Predicts the Future of Integrated Circuits, The Silicon Engine.
3. Intel, Moore’s Law, Intel Newsroom.

Last reviewed 29 October 2026