What is Moore’s Law?
In 1965, Dr. Gordon E. Moore wrote an article for the journal Electronics, wherein he made a famous prediction — that the number of transistors that could fit on a single microchip would double every year for the next decade. This prediction was remarkably accurate, and in 1975, Moore revised his prediction, stating that the number of transistors would double every two years. This new prediction, later
dubbed Moore’s Law by Dr. Carver Mead, has mostly held true into the modern day.
With more transistors, computers can store, process, and transmit more data. This has a myriad of benefits, from faster internet browsing to better video resolution to more accurate machine learning models. Experts in the electronics industry use Moore’s Law to predict economic trends, decide which projects to fund, and plan product development and release dates.
But how is a 50-year-old prediction so accurate? How long will it last? And why does the number of transistors matter to begin with?
Understanding Moore’s Law

The original prediction made in 1965 was simply an extrapolation from the existing trend since 1959, however, Moore did provide some insight into why this trend shouldn’t break. At the time, the most cost-effective number of components per chip was 50, but the theoretical limit was 250 000 per square inch. One potential problem with creating such a densely packed structure was the presence of defects
during the manufacturing process. Moore argued that it was possible to avoid these defects entirely with careful enough engineering processes, but there simply wasn’t a financial incentive to do so at the time. Another potential problem would be heat generated when electricity flows through the transistors. Any time electricity passes through a material, it generates some heat due to resistance (the only exception being zero-resistance materials called superconductors). A large number of transistors
could produce enough heat to damage the device. Thankfully, microchips are extremely thin, therefore there is a large surface area to dissipate heat from, cooling the whole structure down.
When reviewing his prediction in 1975, Moore analysed the many contributions to the improvement in transistor count. First, the microchip size itself had increased. At the time (and to this day), most microchips were made on a single silicon wafer before being cut into several pieces, called die. Silicon is a semiconductor with a uniform structure, but if a defect in the wafer makes it into the die, that microchip will not function properly. For this reason, the die size was kept small to reduce odds of a
defective chip. By 1975, the die size had increased to about a centimetre on each side due to reductions in defect density.
The second contribution is that the size of each transistor and the spacing between them had decreased, resulting in a 32-fold increase in component density. The final contribution was something Moore called “circuit cleverness,” which simply meant arranging transistors in such a way as to minimize the need for inactive components, such as circuit isolation or connectors. By 1975, microchips were as
clever as they could get, thus the rate of growth had begun to slow. Moore removed this contribution from his new prediction, resulting in a doubling of transistors every two years instead of one.
Which leads us to today, where transistor counts are in the tens or even hundreds of billions for commercial use. Looking at more specialized applications, trillions of transistors is not out of reach. Cerebras, a company that trains machine learning models, recently announced the WSE-2, which boasts 2.6 trillion transistors on a single 21.5cm by 21.5 cm die.
So why is the number of transistors so important?
The Importance of Transistors
Computers communicate using binary, a system that is very similar to Morse code. In Morse code, each letter of the alphabet is represented using a unique combination of short and long signals, called dots and dashes. These signals can be audible, like a beeping noise, or visual, like a flashing light. To convey a sentence, the signals for each letter in that sentence are played sequentially. While we could transcribe this signal using periods and dashes, an alternative method is to represent the short signal with a zero and the long signal with a one.
Computers do this with electric current as the signal. When the current is ON, it represents a one, and when the current is OFF, it represents a zero. Acting as a switch, transistors rapidly alternate between these two states, which we call bits. As each letter requires a unique sequence of bits, multiple bits are necessary to transmit the entire alphabet, along with spaces, numbers, and special characters. The Unicode Consortium has established an encoding standard that ensures software companies around the world use the same sequences of binary to represent the same characters in multiple languages. This system uses eight bits or sixteen bits per character.
Unlike Morse code, a few transistors in the right arrangement can change the input signal into a useful output. These arrangements, called logic gates, are the heart of computing. The AND gate for instance, takes an input from two transistors, and gives a single output. Specifically, the output will be one if both inputs are one, and the output will be zero otherwise. Despite their complexity, modern computers
can be broken down into these logic gates, but it wouldn’t be practical for programmers and computer scientists to do so. Instead, they work at much higher abstraction layers, such as machine code or programming languages.
Field Effect Transistors
In order to fit more transistors on a single chip, it’s important to understand how they work and how they control the flow of electricity. The modern transistor is something called a Field Effect Transistor (or FET). Electricity flows from the source to the drain through a piece of semiconducting material, called the channel. Normally, semiconductors are not good conductors, but this one is “doped” with a few
impurities that either give it too many electrons (n-type) or too few (p-type). With these extra charge carriers, the material becomes conductive.
On either side of this material is the gate, which is made of semiconductor with the opposite doping. For example, if the channel is n-doped, the gate will be p-doped. Applying a current or voltage to the gate creates an electric field that either aids or impedes the flow of electricity through the channel.
In the context of computing, it is the voltage applied to the gate that serves as our input, whereas the flow of electricity into the drain is the output. Creating transistors like this allows for very rapid switching between states, thus speeding up processing speeds.
The Future of Moore’s Law
In recent years, the number of transistors per microchip has fallen slightly below the prediction made by Moore’s Law. This is mainly due to the fact that we are running into some fundamental physical limits. For example, transistor channels are only a few nanometres thick, but silicon atoms have a radius of 0.132 nm, so they can’t get much thinner. Furthermore, at such small scales, electrons tend to spontaneously change locations (an effect called quantum tunneling). This isn’t ideal if signals are
sent via electrical current. Finally, the amount of heat generated by these transistors is becoming a problem because there is not enough space between transistors to dissipate it.
However, there is still hope. New technologies, such as finFETs, take advantage of the third dimension for a more efficient use of space. Additionally, even if the number of transistors stays static, more efficient software can make better use of the same hardware.
Summary
Computers store, process, and transmit information using binary that is encoded in electrical signals. This current is rapidly switched on and off using transistors. Moore’s Law was revised in 1975 to predict that the number of transistors per microchip would double every two years, which has held true to the modern day. Although we have begun to fall behind in recent years, technological innovations and careful planning using current technology will allow us to continue making faster and smarter computers.





