Jevons paradox
Making something more efficient often increases its total consumption.
- Making something more efficient often increases its total consumption.
- Operating: Productivity tools that make individual workers faster rarely shrink the total work done.
- Investing: When AI inference costs drop 10x, total AI spending doesn't drop.
- Everyday life: Dishwashers were sold as saving water vs hand-washing.
William Jevons in 1865, writing about coal: "It is wholly a confusion of ideas to suppose that the economical use of fuel is equivalent to a diminished consumption. The very contrary is the truth."
He noticed that as steam engines got more fuel-efficient, total coal use went up, not down. Each engine used less coal per hour, but the efficiency made coal-powered machinery cheaper to run, which made it economic in more places, which drove total coal consumption higher than before.
The pattern repeats endlessly:
- Cars got more fuel-efficient. People drive more miles, and the country has more cars. Total fuel use is roughly flat or up.
- LEDs are 10x more efficient than incandescent bulbs. Lighting expanded so dramatically (decorative lighting, large screens, signage) that total electricity used for lighting actually went up.
- Internet bandwidth got radically cheaper per gigabyte. People didn't use less bandwidth, they used wildly more. Streaming video, 4K, cloud everything. Total bandwidth has roughly tripled every few years.
- Cloud computing made server capacity 10x cheaper than running on-prem. Companies didn't shrink their infrastructure spend. They ran 100x more workloads.
The mechanic: efficiency makes something cheaper to use. Cheaper use opens up applications that weren't economic before. New applications drive total demand higher than the per-unit savings.
For operators, this matters in two ways.
First, don't assume an efficiency win in one area shrinks the total cost line item. If your engineering team finds a way to cut cloud spend by 30%, the team will often respond by running more compute, not by banking the savings. Total spending stays flat or grows.
Second, when evaluating efficiency plays, ask whether the market will absorb the savings as new demand. If yes, the efficiency play is actually a growth play. If no (because the market is genuinely capped), it's a savings play. The two have very different valuations and very different strategic implications.
Jevons paradox is the systems-level cousin of the cobra effect (see [cobra-effect]). Both are about second-order responses to interventions that look obviously good on first inspection.
Examples in the wild
Productivity tools that make individual workers faster rarely shrink the total work done. They usually just expand the volume of work the company can handle. The headcount stays. Email volumes go up.
When AI inference costs drop 10x, total AI spending doesn't drop. It goes up because new use cases open. The shovelmakers (NVIDIA, AWS) make more money even though per-call cost drops.
Dishwashers were sold as saving water vs hand-washing. People who got dishwashers ended up using more total water for dishes because they did dishes more often (smaller loads, more frequent).
Jevons paradox is one of the mental models we apply through real cases inside the Pareto MBA — a part-time program for professionals who want to think clearly about business.