Moore's Law

Moore's law is the observation that the number of transistors that can be placed on an integrated circuit doubles at a regular interval, roughly every two years. It is named after Gordon Moore, a co-founder of Fairchild Semiconductor and later of Intel, who first described the trend in 1965. Although called a law, it is an empirical pattern and an industry target rather than a law of physics.

The steady doubling meant that computing power became dramatically cheaper and more plentiful decade after decade, transforming industries from banking and payments to communications and, later, cryptocurrency. Moore's law is often grouped with other technology rules of thumb, including Metcalfe's law on networks and Ray Kurzweil's broader law of accelerating returns. Since the 2010s the pace of improvement has slowed, and the industry debates whether the law has ended or simply changed form.

Imagine a city where, every two years, builders find a way to fit twice as many homes on each street without making the streets bigger. After a few rounds, each street holds dozens of times more homes than at the start, and the cost of each home falls sharply. Computer chips followed that pattern: engineers kept shrinking the tiny switches, called transistors, so more of them fit on each chip.

More switches meant faster, more capable computers at the same or lower price. That is why a modern phone is far more powerful than the room-sized computers banks used in the 1960s, and why tasks that once needed a data centre can now run in a pocket.

If transistor counts double every two years, the number after $t$ years is:

• $N(t)$ is the number of transistors after $t$ years.
• $N_0$ is the starting number of transistors.
• $t$ is the number of years elapsed.

For example, a chip with 1 billion transistors would, after 10 years of doubling every two years, have 1,000,000,000 × $2^{5}$ = 1,000,000,000 × 32 = 32 billion transistors. Over 20 years the multiple would be $2^{10}$, or 1,024 times.

In a 1965 article in Electronics magazine, Moore noted that the number of components on the most cost-effective chips had been doubling about every year and predicted the trend would continue for at least a decade. In 1975 he revised the rate to a doubling about every two years. The semiconductor industry then adopted the trend as a planning target, coordinating research roadmaps so that manufacturers, equipment makers and software developers could plan around it. For several decades, shrinking transistors also made each one faster and more energy efficient, a related effect known as Dennard scaling, which broke down in the mid-2000s as heat became a limiting factor.

• Banking and payments. Cheaper computing enabled real-time card authorisation, online banking, mobile payments and the cloud computing services that many fintech firms now rely on.
• Trading. Faster processors and networks made electronic markets and high-frequency trading possible.
• Crypto mining. Bitcoin mining moved from processors to graphics cards and then to ASIC chips, and each generation of chips has delivered more hashes per unit of energy. Network hashrate has grown far faster than chip density alone, as both efficiency and investment have risen.
• Cryptography. Rising computing power is one reason security standards keep increasing key lengths, and why systems are designed with large margins against brute-force attack. Quantum computing is a separate threat that does not follow Moore's law.

Transistors are now so small that physical limits, such as the size of atoms, current leakage and the heat produced in dense chips, make further shrinking harder and more expensive. The time between new manufacturing generations has lengthened, and the cost of building leading-edge factories has soared, concentrating production in a few companies. Many engineers say Moore's law in its original form has slowed or ended, while others point to new techniques, such as three-dimensional chip stacking, specialised processors and advanced packaging, as ways of continuing the gains in performance.

Economists also caution against treating the law as a general rule for technology. It describes one industry's unusually steady progress, partly made self-fulfilling by being used as a target. Other technologies, such as batteries or engines, have improved much more slowly, and extrapolating exponential growth from one field to all of them can lead to over-optimistic forecasts. The wider link between computing advances and economic growth is debated in research on productivity, which has not always risen as fast as computing power might suggest.