How Technology Compounds

Most of the tools we use every day did not exist fifty years ago, and many did not exist ten years ago. That pace is not an accident. Technology tends to build on itself, so each advance makes the next one cheaper and faster to reach. This essay looks at how that compounding works, the best-known evidence for it, and the forces that push back.

New technology is made from old technology

The economist W. Brian Arthur argues in The Nature of Technology (2009) that technologies are built by combining earlier technologies. A smartphone is a radio, a computer, a camera, a battery, a touchscreen and a satellite receiver put into one case. Each of those parts is itself a combination of older parts.

If that is true, the number of possible new combinations grows as the stock of existing technologies grows. Every invention adds a new building block for the next round. This is the simplest explanation for why progress can speed up instead of running at a steady rate.

Moore's Law

The clearest example is the integrated circuit. In April 1965, Gordon Moore, then at Fairchild Semiconductor and later a co-founder of Intel, wrote a short article in Electronics magazine titled "Cramming more components onto integrated circuits." He observed that the number of components on a chip had been doubling about every year and expected it to continue. In 1975 he revised the pace to a doubling about every two years.

The prediction became a target the whole industry planned around. Intel's 4004 processor in 1971 held about 2,300 transistors. Nvidia's Blackwell graphics processor, announced in 2024, holds 208 billion. That is an increase of nearly 100 million times in about fifty years.

Moore's Law is not a law of nature. It describes what engineers, manufacturers and investors chose to achieve, and it required enormous spending at every step. Since the mid-2000s, some of its benefits have slowed. Chips stopped getting faster clock speeds around 2005, and the cost of each new manufacturing plant has climbed into the tens of billions of dollars. Progress continued anyway through more processor cores, specialized chips and new ways of stacking and packaging silicon.

The same curve in other fields

Computing is not the only field with a steep cost curve. The U.S. National Human Genome Research Institute has tracked the cost of sequencing a human genome since 2001. It started at about $100 million and fell to around $1,000 within two decades, faster than chip costs fell over the same years. Cheaper sequencing then made new medicine and new research possible, which is compounding at work.

Energy shows a similar pattern. The price of solar photovoltaic modules has fallen by more than 90 percent since 2010, according to the International Renewable Energy Agency. Lithium-ion battery pack prices have followed a similar path. Each drop opened new markets, and each new market paid for the next round of improvement.

A common thread runs through these examples. When a technology gets cheaper, more people use it. More use brings more money and more learning, and those make the technology cheaper again. Economists call the pattern a learning curve or experience curve. It was first described for aircraft production by Theodore Wright in 1936.

Why it takes longer than it looks

Fast improvement in a technology does not mean fast change in daily life. In 1990 the economic historian Paul David published "The Dynamo and the Computer." He showed that the electric motor was available in the 1880s, but factories took about forty years to gain much from it. To get the benefit, owners had to redesign whole buildings around small motors at each machine, instead of one big shaft driven by a steam engine. Workers had to learn new methods too.

David used that history to explain why computers were everywhere in the 1980s but barely showed up in productivity statistics. The gain came later, in the late 1990s, once businesses had reorganized around them. The lesson is that a powerful new tool often needs years of changes in organizations, skills and rules before it delivers.

Waves, not a straight line

Carlota Perez, in Technological Revolutions and Financial Capital (2002), describes five great surges since the late 1700s: the industrial revolution, steam and railways, steel and heavy engineering, oil and mass production, and information and telecommunications. Each one followed a similar pattern: an installation period of excitement and speculation, often ending in a financial crash, then a longer deployment period in which the technology spread through the whole economy.

Her model is a reminder that compounding is not smooth. Progress arrives in bursts, and the bursts are shaped by money, policy and public trust as much as by engineering.

What slows it down

• Physics. Transistors are now only a few dozen atoms across in places. Some limits cannot be engineered away.
• Cost. Each new generation of chips, drugs or reactors can cost more to develop than the last.
• Complexity. Larger systems are harder to understand, test and maintain.
• People and rules. Adoption depends on skills, regulation, trust and the will to change how work is done.

None of these forces has stopped technology from compounding so far, but each one bends the curve. The question for any field is which force wins over the next decade, and the answer changes from field to field.

Why it matters

Compounding is easy to underestimate. A process that doubles every two years looks slow for a long time and then becomes very large very quickly. Much of today's debate about artificial intelligence, covered in the next essay, is a debate about how long that kind of curve can continue.

Next: A Short History of Artificial Intelligence