Moore's law

In 1965 Gordon Moore, a co-founder of Intel, noticed a striking trend and predicted it would continue: the number of transistors that engineers could pack onto a chip was roughly doubling about every two years, as each transistor shrank. For decades this prediction held with uncanny steadiness, and it became the metronome of the whole industry — chips got more crowded with switches, generation after generation, almost on schedule.

It is important to be precise about what the law is and is not. Moore's law is about transistor count — how many switches fit on a chip — not directly about speed. It is an economic and engineering observation, not a law of physics; it held because the industry organised itself around delivering it. More transistors gave designers the raw material to build bigger caches, more cores, and cleverer microarchitectures, which is how the transistor bounty turned into useful capability.

Be honest about today: Moore's law has dramatically slowed and, by many measures, is effectively ending, because transistors are now only a handful of atoms wide and shrinking further is brutally hard and expensive. Crucially, even while transistor counts grew, the companion trend that turned more transistors into proportionally faster, cooler chips — Dennard scaling — broke down around 2005. So the comforting belief that computers just keep getting faster on their own is no longer true. That ending is precisely why architecture matters more now: with no free ride from shrinking transistors, cleverer design is how we keep improving.