Jevons’ Paradox and the Silicon Roadmap

By Moshe Zalcberg

November 2, 2026

AI usage may expand with more efficient models and hardware, just as Jevons would predict. But the silicon roadmap adds a layer of complexity to the market prediction.

Gartner vs Jevons

A couple of weeks ago,  Gartner published a forecast that raises concerns about AI infrastructure. AI providers will spend close to $1 trillion on infrastructure in 2026 and another $6 trillion between 2027 and 2030. To pay for that, cover the cost of borrowing, and still earn a 12% return, the owners of AI data centers would need about $8.5 trillion in revenue through 2030. Gartner expects the AI revenue that actually reaches those data centers to be $3.6 trillion at most. On that arithmetic, says Gartner, more than $3.5 trillion of this spending could be exposed to write-downs by 2030, and more than 80% of neoclouds will no longer exist as independent companies. That is a risk estimate under Gartner’s assumptions, not a prediction that the full amount will be written off.

On the other hand, IMHO, this math assumes that AI will continue to be deployed in a similar use and business model as today. However, we might well be only at the beginning of the AI use curve. And, according to the by-now very well-known Jevons paradox, the more a technology is available, the more people will find new use-models for that technology, and therefore, even if the technology price goes down, the revenue from the aggregated usage can go up. As an illustration, when cars become more fuel-efficient, each kilometer costs less, so people drive more, take longer trips, and may end up buying more fuel in total. Cheaper petrol works in a similar way: When its price falls, people drive more, and when it rises, some prefer public transportation.

To read the full article, click here