When technology makes a resource more efficient to use, the effective cost of using it falls. This often increases demand. For example, more fuel-efficient cars make each kilometer cheaper, which can encourage people to drive more. As a result, some of the expected reduction in fuel consumption is offset by increased usage. This is known as the rebound effect.
If the increase in demand is large enough to more than offset the efficiency gains, total resource consumption rises rather than falls. This effect is known as the Jevons paradox.
Conditions for the Jevons Paradox
Three conditions generally need to be met for the Jevons paradox to occur:
- Technology increases efficiency or productivity, allowing the same output to be produced with fewer resources.
- These efficiency gains reduce the effective cost of the resulting product or service.
- Demand responds strongly to the lower cost, causing consumption to increase enough to offset the original efficiency gains.
The Jevons paradox occurs when this increase in demand is so large that total resource consumption ultimately rises despite the improvement in efficiency.
AI and the Jevons Paradox
Microsoft CEO Satya Nadella has referenced the Jevons paradox when describing artificial intelligence as potentially becoming a commodity that we simply cannot get enough of.
Erik Brynjolfsson, an economist at Stanford University who studies the impact of digital technologies and AI on productivity and work, argues that some occupations may already be experiencing a Jevons-paradox-like effect. Software developers and radiologists, for example, are becoming more productive through AI, yet demand for their work may continue to grow rather than decline.
Increased productivity can lower the cost of performing these tasks, potentially expanding their use and creating additional demand for the people who perform them.
There are already indications that the overall volume of software development continues to expand. GitHub reported more than 121 million new repositories in 2025, its largest annual increase to date, alongside record levels of developer activity. While repository and contribution counts are imperfect proxies for the amount of software being produced, they suggest that software creation is expanding rather than contracting.
Custom-Developed Solutions Become Affordable
Historically, many custom software solutions were simply too expensive to build. Organizations either relied on off-the-shelf software, adapted their processes to existing tools, or did not develop a solution at all.
AI changes this economic equation. As AI increases developer productivity, the cost of building and maintaining software can decrease. This makes custom development economically viable for a much larger number of use cases.
I therefore expect AI to lead to more, not less, custom-developed software. Instead of replacing developers simply because each developer can produce more software, higher productivity may expand the overall demand for software by making tailored solutions affordable in situations where they previously were not.
Rebound Effect or Jevons Paradox?
Whether we will actually see a Jevons effect in software development remains to be seen. Suppose AI makes developers twice as productive. If this results in three times as much software being demanded, the overall demand for development work would increase despite the productivity gain.
If, however, doubling productivity leads to only a 50 percent increase in demand for software, we would still see a rebound effect, but not a full Jevons paradox.
The distinction matters. AI may allow us to build software with fewer developers per application while simultaneously making software development attractive for many more problems. Whether this ultimately reduces or increases the total demand for developers depends on which of these effects is stronger.