What is the Jevons paradox?

The Jevons paradox is the effect by which making a resource more efficient to use raises its total consumption instead of lowering it. The naive intuition is that if a unit of that resource costs less or delivers more, total spending should fall for the same need. The opposite happens when the lower unit cost opens up uses that were not worth it before: demand grows faster than the unit cost falls, and the total goes up. A budget planner can get the arithmetic right and still miss the forecast, because a market dynamic is at work: the cheaper an input becomes, the more reasons there are to use it. For anyone managing capacity, cloud, or AI spend, the paradox is why "the cost per unit is down by half" and "the bill went up" can both be true in the same quarter.

See also: FinOps · The practice bringing financial accountability to the cloud: every team sees, understands and optimizes the cost of what it runs.

Why efficiency expands demand

The name comes from William Stanley Jevons, who in the essay The Coal Question (1865) observed that James Watt's steam engine, far more efficient in its coal use than earlier engines, did not reduce Britain's coal consumption: it sent it soaring, because it made installing steam engines worthwhile in places where they had never paid off before. The mechanism is the price elasticity of demand, measured against the effective price: if a cost drop opens up marginal use cases that were previously ruled out on cost grounds, the usage base grows by more than the unit cost shrank. It does not always happen: if demand is already saturated or inelastic, an efficiency gain does translate into real savings. Before promising savings, then, estimate how many new uses become worthwhile at this price, which is also the calculation to redo in a business case whenever an input's unit cost changes, even if the project itself stays the same.

An enterprise example

The most cited AI case is Satya Nadella's: in January 2025, after the release of the Chinese model DeepSeek and the collapse in inference costs that followed, he explicitly invoked the Jevons paradox to explain why Microsoft did not see that price drop as a threat to compute demand. In Microsoft's second-quarter fiscal 2025 earnings call he said: "When token prices fall, inference computing prices fall, that means people can consume more". The same logic shows up at smaller enterprise scale: a company that negotiates a lower per-token price often does not cut its AI spend, it redirects it toward use cases that did not previously justify the cost (assistants on every internal process instead of just one, agents that repeat the same call several times to check the answer), and the monthly bill stays flat or grows.

Why it matters for decision makers

The Jevons paradox is not an argument against efficiency: it is a warning against the wrong kind of forecast. Anyone who plans an AI or cloud budget on unit price alone, without estimating how much that lower price will grow volume, is reliably surprised by the bill. You keep optimizing unit cost and you govern the volume that optimization unlocks: budgets with thresholds, visibility into who consumes what and for which use case, explicit criteria for which new uses are worth enabling at the new price. That is the ground FinOps covers, since FinOps is what measures and governs that volume in practice, and the ground of AI tokenomics, where the drop in cost per million tokens is exactly the kind of efficiency gain the paradox describes.