General Purpose Technology

Some inventions stay small. The zipper is useful, but it only changes clothing and bags. Other inventions reach into everything: electricity powers lamps, machines, trains, refrigerators, and data centers. Economists call such techniques General Purpose Technology, roughly translated as foundational technology. What is meant is a technique that can be used across a great many fields, that improves substantially over decades, and that in doing so makes other inventions possible in the first place. Classic examples include the steam engine, electricity, and the computer.

Why economists apply this label so sparingly

The term is not a marketing phrase but a fairly high hurdle. For a technique to count as a General Purpose Technology, three things must come together. It must be broadly applicable, it must become significantly better and cheaper over a long period, and it must trigger new inventions in other fields. Most innovations meet at most one of these conditions. Historians therefore count, depending on how one counts, only about two dozen such techniques in the entire history of humanity.

This classification matters because it says something about time and money. Foundational technologies do not immediately make an economy more productive. At first, the transformation only costs money: new machines, new buildings, new training. The economist Paul David demonstrated this with the electrification of US factories. Electric motors already existed in the 1880s. But the great productivity boost only came around 1920, when factories were finally built differently.

This leads to an uncomfortable expectation for investors and companies. A genuine foundational technology can disappoint for years and still end up changing everything. Conversely, early enthusiasm on the stock market does not mean that such a technology is actually present.

The long road from invention to transformation

The sequence tends to be remarkably similar each time. At the start there is a core technique that is still expensive and unreliable. Then costs fall, because many providers work on it and unit numbers rise. Only then does it become worthwhile to use the technique in areas nobody originally thought of.

The decisive step consists of what are called complementary investments — everything additionally required. In the case of electricity, that meant power grids, power plants, standardized plugs, and rebuilt factory halls. In the case of the computer, it meant software, networks, and employees who knew how to use them. These accompanying investments usually cost a multiple of the actual technology itself. They are also the reason why the benefit becomes visible so late.

A common mistake is simply rebuilding the old way of working with the new tool. Early factories replaced the steam engine with one large electric motor and left the drive shafts under the ceiling where they had always been. The gain was small. Only once every machine got its own small motor could the hall be arranged according to the actual workflow. The technical term for this is the productivity paradox: the technology has long since arrived, but the statistics do not yet show it.

The debate over AI as the next foundational technology

In business news, the term currently appears almost always in connection with Artificial Intelligence. Proponents argue that language models are useful in medicine, law, programming, administration, and research all at once. This very breadth is the strongest argument. In addition, the cost per query has fallen dramatically since 2022.

Skeptics counter that the measurable effect on the productivity of entire economies has so far been small. They point to earlier cases such as nanotechnology, which people had expected similar things from. Whether AI clears the hurdle can only be answered seriously in hindsight, presumably in twenty years.

For you as a reader, the term is above all a warning sign about language. When a company describes its own technology as a General Purpose Technology, that is a claim about the future, not a finding. The sensible question is whether industries that originally had nothing to do with the technology are already being transformed by it. That is exactly what distinguishes a foundational technology from a very successful product.