AI and the Pitfalls of Innovation
Paul Krugman examines how historical examples such as electrification, the postwar boom, and the IT productivity paradox can help understand AI's economic impact. He argues AI is not a passing fad but its short- and long-term effects remain uncertain, and he plans a series of articles on the topic.
Paul Krugman
Jun 07, 2026
∙ Paid
AI is certainly not a passing fad. However, nobody knows how it will affect the economy. In the short run, there’s much room for debate about whether the rush to build datacenters and AI-ify everything is a bubble. And in the long run, there’s even more scope for argument about the impacts on productivity, employment and wages.
So many people, myself included, are looking for historical examples that may provide guidance on how AI will affect the economy. Granted, quizzing history for insight into the effects of a radical innovation is somewhat odd: By definition, a transformative new technology has never before been actualized. So how can the past teach us about its effects? Still, as a motto often (but without evidence) attributed to Mark Twain puts it, history may not repeat itself, but it rhymes. While AI is something entirely new, over the past two centuries there have been many introductions of radical new technologies. So an investigation of these episodes may provide valuable insights for the future.
When one uses history to make sense of the present, however, it’s important to have a wide view. A number of smart observers, including Azeem Azhar and John Burn-Murdoch, have been leaning hard on a classic example of radical technological change that took a long time to fully bear fruit: electrification in the late 19th and early 20th century. That’s a good choice, because studying that example helped economists understand and predict the delayed payoff to the rise of modern information technology (IT).
Yet there are other episodes that I believe deserve to be given equal weight: The great postwar productivity boom, which is notable because it wasn’t driven by radical new technologies, as well as the disappointingly early petering out of the IT-driven productivity boom of the 1990s and 2000s.
Today’s primer will be the first of what I expect to be a multi-part series on the economics of AI. Today I will focus on the history of productivity, while reserving extended discussion of AI’s future, fears of technological unemployment, effects on income distribution and more for subsequent posts.
Beyond the paywall I will address the following:
- How economists measure the impact of technology
- The mystery of the great Post-war boom
- The Solow paradox: Why was the payoff to IT so slow to arrive?
- The IT disappointment: Only 10 years of productivity payoff?
- AI: Preliminary questions
Continue reading this post for free, courtesy of Paul Krugman.
Or purchase a paid subscription.