01What is Moore's Law?
In 1965, Intel co-founder Gordon Moore initially stated that the number of components on an integrated circuit would double in about 18 months. In 1975, he changed the law to mean that the number of transistors on a chip per area would double every two years. Moore's Law has been driving the rapid development of integrated circuits and chip industries ever since.
Is Moore's Law true or not? Let's take a look at the graph below, where we can see that the sampling points are basically near the curve, and we can see that Moore's Law is still basically accurate.
Moore's Law is not a mathematical or physical law, but a prediction of trends, so there should be some margin for both verbal and quantitative calculations. In this sense, Moore's prediction is quite accurate, so it will be recognized by people in the industry, and have a huge response.
02Why Is Moore's Law in Trouble?
The problem with Moore's Law in 2022 is that transistors are now so small that we can't do much to make them smaller. According to TSMC's 2024 production roadmap, the width of the transistor gate (the part of the transistor through which electrons flow as a current) is now close to 2 nanometers.
A silicon atom is 0.2 nanometers wide, which makes a 2-nanometer gate about 10 silicon atoms long. At these scales, it becomes increasingly difficult to control the flow of electrons as various quantum effects come into play within the transistor itself. For larger transistors, crystal deformations at the atomic scale don't affect the overall flow of current, but when you're only about 10 atoms away to work with, any changes in the underlying atomic structure will affect the flow of current through the transistor.
Another potential pitfall of Moore's Law is simple economics. The cost of shrinking transistors has not fallen as much as it did in the 1960s. When demand for semiconductor chips took off, the engineering capacity to produce them was expensive, but at least available. As demand for everything from smartphones to satellites to the Internet of things soars, there is not enough capacity to meet it, causing prices to rise at every step of the supply chain. What's more, as the number of transistors doubles, so does the heat they generate, making it increasingly unaffordable to cool large server rooms.
As companies try to extend the life and performance of their existing equipment to save money, the chipmakers responsible for implementing Moore's Law bring in less revenue for research and development -- and research itself becomes more expensive. Without additional benefits, it would be harder to overcome all the physical obstacles to further shrinking transistors. So even if the physical challenges won't end Moore's Law, the lack of demand for smaller transistors will almost certainly.
03Moore's Law is Dead! Long live Moore's Law
Ultimately, Moore's Law was never a "law" in the first place, but more of a self-fulfilling wish. We expect transistor density to double every year, and then every two years, so we looked at how to do that
Whatever comes next, be it quantum computing, machine learning and artificial intelligence, or even something we don't even have a name for yet, we'll find a new aspiration to drive this innovation forward. Moore's Law may have run its course, but if we want it bad enough, we'll find a new Moore's Law.