Moore's Law

A computer chip is a piece of silicon covered with tiny electronic switches. These switches are called transistors, and the more of them fit on a chip, the more it can compute. Engineer Gordon Moore observed in 1965 that their number doubled at a regular pace. Later he revised the rhythm to about two years. This observation is called Moore’s Law. It is not a law of physics but a prediction that has held true for a remarkably long time.

Why the doubling changed so much

A doubling every two years sounds harmless. But over twenty years it amounts to a factor of a thousand. This exact pace explains why a smartphone today outperforms a mainframe computer from the 1980s. The first Intel processor from 1971 had around 2,300 transistors. Modern chips have several dozen billion.

For the industry, this was more than just an observation. Chip manufacturers planned their factories around it, and software companies calculated their plans based on future faster hardware. You could write programs that still ran too slowly on current computers. Two years later, they became usable. Moore’s Law thus became a self-fulfilling prophecy: everyone aligned their plans with it, so it came true.

The boom in artificial intelligence also depends on it. Large language models require enormous computing power. This computing power was only affordable because chips became ever more powerful and cheaper per computing step over the decades.

What happens as transistors shrink

More transistors on the same area means each individual one has to become smaller. The structures on today’s chips are only a few nanometers wide. A nanometer is one millionth of a millimeter. For comparison: a human hair is about 50,000 nanometers thick. These structures are manufactured using light that projects a pattern onto the silicon wafer, similar to a slide projector.

Smaller transistors have two advantages. They need less power, and signals have to travel shorter distances. This makes the chip both faster and more energy-efficient at the same time. This relationship used to be called Dennard scaling. It has largely collapsed since around 2005, because very small transistors become disproportionately hot.

Since then, processor clock speeds have barely increased. Instead, manufacturers build multiple computing cores side by side or stack chips on top of each other. A common misconception is therefore to equate Moore’s Law with speed. It only says something about the number of transistors, not about speed.

The term in stock market news and product announcements

In business news, Moore’s Law usually comes up with the question of whether it still holds. Some experts declare it dead because the physical limits are close. Others point out that the transistor count keeps rising, just more slowly and at significantly higher cost. A modern chip factory now costs more than 20 billion dollars.

You’ll encounter the term specifically when companies like TSMC, Intel, or Samsung introduce new manufacturing processes. Labels like “3 nanometers” have long since become marketing names rather than actual measurements. Nvidia's graphics cards for AI data centers are also often advertised as surpassing the old rhythm.

In everyday life, you notice the consequences in that new phones or laptops are no longer noticeably faster every year. Progress is shifting toward specialized chips that excel at just one task. One example are the AI accelerators in current smartphones.