LIFE IS A VOYAGE ARE YOU PREPARED?

The Data Centers Eating the Power Grid

Let’s start with the most concrete and measurable consequence of AI’s explosive growth: electricity.

Training a single large AI model consumes roughly the same amount of energy as the lifetime carbon footprint of five average American cars. That was considered shocking when researchers first published that figure a few years ago. Today it barely registers as a footnote, because the scale of what has been built since then makes those early models look like pocket calculators.

Data centers the physical infrastructure that powers every AI query, every chatbot conversation, every image generated and email drafted by machine are now among the fastest growing consumers of electricity on the planet. In the United States alone, data center electricity demand is projected to more than double by 2030. In some regions, power grid operators are openly warning that existing infrastructure simply cannot keep up.

Here is where Jevons arrives, uninvited but entirely predictable.

The AI industry’s response to rising energy costs has been exactly what you would expect: build more efficient chips, optimize models, reduce the energy required per query. And they have succeeded. Today’s AI models are dramatically more efficient than their predecessors. A query that required significant computing power two years ago can now be handled at a fraction of the energy cost.

So energy consumption is falling, right?

Not even close. Because as each query got cheaper, the number of queries exploded. Businesses that once ran a handful of AI tasks per day now run thousands. Consumers who once typed a Google search now have extended conversations with AI assistants that each require orders of magnitude more compute than a simple search ever did. Every efficiency gain was immediately consumed and then some by a surge in new demand.

Jevons is not smiling. He is nodding.

The Jobs Paradox Nobody Is Talking About

The energy story is dramatic, but the Jevons Paradox has an equally uncomfortable application to the labor market and this one hits closer to home for most people.

The dominant narrative around AI and jobs goes something like this: AI will automate certain tasks, workers will be displaced, but new jobs will eventually emerge just as they did during the Industrial Revolution. It is a reassuring story. It is also incomplete.

Here is the Jevons version of the jobs story.

When AI makes a worker more productive, the natural assumption is that you need fewer workers to do the same amount of work. A law firm that once needed ten associates to review contracts can now do it with three, assisted by AI. A marketing team that needed six copywriters can now produce the same volume with two. Efficiency goes up. Headcount goes down.

But Jevons asks: what happens to the demand for legal work, marketing content, financial analysis, and software code when the cost of producing it falls dramatically?

It expands. Possibly without limit.

Small businesses that could never afford a lawyer can now access AI-assisted legal guidance. Startups that had no marketing budget can now produce professional content at scale. Companies that once built simple software products because complex ones were too expensive can now build sophisticated applications with lean engineering teams. The total demand for legal work, marketing, code, and analysis does not shrink when AI makes it cheaper. It grows often explosively.

The uncomfortable implication is that AI may not simply replace jobs. It may restructure entire industries around a new cost curve one where the volume of work expands so dramatically that the workforce required looks nothing like what came before. Some workers will thrive. Others will find that their specific skills have been commoditized in ways that no amount of retraining fully addresses.

The workers most at risk are not necessarily those doing the simplest tasks. They are those doing high-volume, repeatable cognitive work the kind of work that feels skilled and specialized but follows predictable enough patterns that AI can learn to replicate it. Think junior analysts, entry-level coders, paralegals, and mid-level content creators. The Jevons Paradox does not eliminate the need for their work. It eliminates the need for them to do it.

The Biggest “Ask Jevons” Question

There is a version of the AI story that is genuinely optimistic. Cheaper intelligence means more people have access to expertise that was once reserved for the wealthy. A first-generation college student in rural America can access the same quality of financial planning guidance, legal advice, and medical information as someone in a major metropolitan area. That is not nothing. That is potentially transformative.

But Jevons asks us to hold that optimism honestly. Every technology that promised to make life easier, cheaper, and more efficient has also made us consume more; more energy, more resources, more of everything. The printing press did not reduce the demand for information. The automobile did not reduce the demand for travel. The internet did not reduce the demand for human attention. It expanded all of it, in ways that created enormous wealth and enormous problems simultaneously.

AI will be no different.

The question is not whether the Jevons Paradox applies to artificial intelligence. It already does, visibly and measurably, in the data centers humming across America, in the power purchase agreements being signed between tech giants and utilities, in the labor market data that is only beginning to reflect the scale of the transition underway.

The question is whether we are paying attention.

Ask Jeeves is gone. But the questions he represents who has access to knowledge, what it costs to answer them, and what we do with the answers have never been more alive.

Welcome to the age of Ask Jevons.