Jevons-paradoxon

in English: Jevons paradox

advancedwidespread

If we can use something more efficiently and cheaply, we often do not consume less of it, but more overall, because we start using it for many more things.

In plain terms

Common sense would lead you to expect that if a car consumes half as much gasoline, then half as much gasoline is consumed. In reality, because driving is cheaper, we drive more and further, and people who did not drive before get into cars, so part of the savings is lost. In the case of cars, measurements show that this is mostly only a partial loss. However, if consumption ultimately becomes greater than before the innovation – as happened with coal after more efficient steam engines in the 19th century – it is called the Jevons paradox.

The same question applies to AI. Producing answers of the same quality becomes cheaper and requires less energy every year. Yet total computing demand and electricity consumption do not decline, because we use AI for far more tasks: search, customer service, programming, and agents working independently for hours.

The concept also comes up in debates about work. If AI makes, for example, software development or translation cheaper, the result may not be less of that work, but so many more orders that the profession survives or even grows. However, this is a possibility, not a certainty.

How it works

A technical innovation makes resource use more efficient: achieving the same result requires less coal, electricity, computing capacity, or labor hours.
This reduces the unit cost of the output: a kilometer traveled, a room lit, or an AI response becomes cheaper.
Lower prices create new demand: existing users consume more, and use cases emerge that were previously not cost-effective.
If demand grows faster than the savings from efficiency, total resource consumption rises. This is called the Jevons paradox.
If demand increases only slightly, part of the savings remains. In this case, we speak only of a partial rebound effect, not a paradox.

An everyday example

An online store initially had AI answer only the most important customer emails, because every request visibly cost money. When the price fell to a fraction, it also handed product descriptions, translation, search and an order-handling agent over to AI. Each request is much cheaper today, yet the company uses far more computing overall, and its monthly AI bill is higher than before.

Who is affected

employeeIf AI makes your work cheaper, demand for it may also rise, but this varies by profession, so it is no guarantee for your job. → Work & HR

entrepreneur, executiveWhile the unit cost of AI drops, your company's total AI bill can easily rise as you integrate it into more processes and agents.

bank customer, investorInvestment arguments around chipmakers, data centers, and energy companies are largely based on the idea that cheaper AI brings more demand. → Money

everyoneDespite more efficient models, AI's electricity and water demand may rise, affecting energy prices and climate goals.

A common misconception

MythThe Jevons paradox is a natural law: efficiency always leads to more consumption, so AI will certainly create more work than it eliminates.

In realityThis only happens if demand reacts very sensitively to price drops. In many cases, only part of the savings is lost, and in some cases demand simply saturates.

Limits and risks

It is difficult to predict in advance whether it will occur in a given field, as it is usually only visible afterward from data.
It is often invoked out of self-interest: the argument "the cheaper it gets, the more of it we need" suits anyone who wants to sell chips, data centers or shares.
Even if demand for a job rises, it is not certain that the same people will perform it under the same conditions and for the same pay.
Demand does not grow infinitely: after a while, the market saturates, and efficiency truly brings savings.

Similar, but not the same

rebound effect – The rebound effect is the broader phenomenon where part of the savings is lost; the Jevons paradox is its extreme case, where consumption becomes even higher than before.
Induced demand – Induced demand arises from increased supply (the new lane attracts more cars), while the Jevons paradox stems from improved efficiency.
Productivity paradox – The productivity paradox means the impact of new technology is not visible in statistics, while the Jevons paradox means that efficiency increases consumption.

Related concepts

Broader

Productivity, Automation, Supply and demand

Narrower

–

On the edge

rebound effect, Induced demand, computing capacity, Data center, AI energy demand

A brief history

English economist William Stanley Jevons wrote in his 1865 book The Coal Question that more efficient steam engines did not reduce but increased coal consumption in England. The concept entered AI debates widely in January 2025, when Chinese company DeepSeek's low-cost model shook the stock market, and Satya Nadella, CEO of Microsoft, cited this paradox.

Where things stand

As of late 2026, it is one of the most frequently cited concepts in economic and investment debates on AI, especially regarding computing capacity and energy demand. However, there is no consensus on how far it applies to AI, and particularly to the labor market.

Status: 2026 November

Also known as

Jevons-paradoxon · Jevons paradoxona · Jevons paradox · Jevons paradox · Jevons' paradox · Jevons effect

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.