Now look at where people are spending once the output turns cheap. In 2025, a record 159 million fans went to a Live Nation show, revenue crossed $25 billion, and for the first time more of them came from outside the United States than inside it. As machine-made output slid toward free, the price of being in a room where something happens once kept climbing.

These two facts look unrelated. They are the same fact. AI didn’t abolish scarcity; it moved it — down into the physical inputs it consumes, and up into the human moments it can’t reproduce. Neither shows up on the dashboards most companies use to track what AI makes cheaper. Most executive teams are asking the right operational questions: where can AI cut cost and accelerate output? Useful questions. But they hide a harder one. When everything you produce becomes abundant, what becomes scarce, and do you still own any of it?

The clearest early picture of the mechanism comes from a corner of the labor market that felt the shift first. When ChatGPT arrived, the people who should have been safest were among the first to feel it. A study of more than three million postings on a global freelancing platform found that within months, demand for the tasks the technology does well fell sharply: translation into Western European languages dropped about 30%, and the most commodified writing, like “About Us” pages, fell by half. A separate analysis of the same market found that the most experienced, highest-priced freelancers were hit at least as hard as everyone else. Skill was no shield.

How can this be explained? Their work hadn’t gotten worse. It had gotten abundant. And abundance, not incompetence, is what destroyed its price.

Abundance is a smokescreen

The abundance AI produces is both real and misleading. What becomes abundant is the output itself, not the conditions that make it usable. Producing an answer has never been easier. Producing the right one, and being accountable when it’s wrong, is still hard.

The freelance market shows the mechanism in miniature. AI didn’t only take work. It devalued the work still being done. The scarce ingredient was never “competent text.” Competent text is now close to free. What stayed scarce was what competent text used to stand in for: judgment, and the assurance that a person had put their name behind the words.

Scarcity goes into hiding

In the industrial economy, scarcity was visible and easy to count: inventory, seats, engineer-hours. In the AI economy it turns diffuse, sometimes invisible. It hides where dashboards don’t look: in a compute queue behind what looks like plenty of stock, or in the seconds of latency that reveal how few machines are actually free. Sometimes the real constraint isn’t technical at all, but legal, or just the sliver of attention a customer has left. These are easy to miss until they bite. A streaming service can surface a thousand titles in an instant, but what it is really rationing is the viewer’s evening, the few minutes anyone will actually spend choosing. A logistics platform prices every route in milliseconds, then runs into a ceiling that has nothing to do with computing power: the line of trucks idling at a single loading dock.

Herbert Simon named the shift back in 1971: an abundance of information produces a scarcity of attention. What was then an insight is now the operating condition of entire sectors. Scarcity doesn’t disappear. It goes into hiding.

Many organizations still reason as if scarcity had been abolished, even as they absorb its effects: congestion, frustrated customers, trade-offs nobody ever chose out loud. They keep measuring units sold and headcount, while the business quietly bleeds through the scarcities they don’t.

Who owns the new scarcity?

The flooded freelance market makes one thing obvious: anyone producing cheap output at scale is renting the scarcity from upstream. The same holds for companies. Drive your marginal cost toward zero and you have usually just pushed the bottleneck one rung up the chain, to whoever controls the compute and the power it runs on.

Data-center electricity demand jumped 17% in 2025 and is on track to roughly double, from about 485 terawatt-hours that year to some 950 by 2030, with demand from AI-focused sites tripling. The capital spending of the five largest technology companies passed $400 billion in 2025 and is expected to climb another 75% this year. By the end of the decade, the IEA expects data centers in the United States to consume more electricity than the production of aluminum, steel, cement, and chemicals combined.

Look at where the constraints now bind, and it isn’t talent or capital. It’s grid connections, gas turbines, transformers, even gallium, of which China refines about 99%. Water belongs on that list now too. The reservoirs draining in Texas are the leading edge, not the exception: direct consumption by U.S. data centers is projected to double or more by 2028, and much of the new build is landing in basins that were already short. The IEA notes these bottlenecks are already capping its more aggressive growth scenarios. The company congratulating itself on near-free production has, in many cases, simply handed the durable margin to whoever owns the scarce input. There is, as the agency’s director likes to say, no AI without energy.

The same logic runs on the demand side. When everyone can generate the same offer, what becomes scarce is where the customer looks: the channel, or the assistant that filters the options. Own that point of attention and you own the scarcity that counts. Everyone else competes over the abundance.

Price no longer says “how much is left.” It says what matters.

When abundance becomes the norm, price changes function. It stops indicating how much is available and starts signaling what is prioritized.

A premium price no longer reflects only a higher cost. It reveals an underlying scarcity: faster access, or less exposure to risk. It doesn’t buy more volume; it buys time and peace of mind. A price set too low is no longer just a margin problem. It can trigger a rush that overloads the system and buries trade-offs that should have been made in the open.

Price becomes the instrument through which a company recognizes a scarcity and decides whether to protect it or charge for it. That is a strategic act, not an arithmetic one. And because AI can measure once-fuzzy constraints in real time, it makes that choice sharper and its consequences immediate.

The companies that grasp this are charging for things nobody thought to sell five years ago. A cloud provider sells not just capacity but guaranteed priority when demand spikes. A financial firm has stopped selling the data, now abundant, and started selling the traceability of its origin and the certainty it will hold up in front of a regulator. In construction software, Graitec has built its AI strategy around the same premise, betting that what an engineer will pay for is not faster generation but a design auditable enough to sign their name to. A wealth manager charges less for the portfolio, an allocation a robo-advisor now assembles for a few basis points, and more for the judgment that keeps a client invested through a crash, the coaching Vanguard advisor’s alpha adds more value than stock selection and the one thing no model will answer for. In each case the price stops rewarding the effort of production and starts putting a number on a scarcity the customer cannot find anywhere else.

The question few executives ask

The leaders handling this transition best share one habit. They don’t only ask what AI lets them produce more of. They can say plainly what, in their business, has just tipped into abundance, and what has just become rare.

Three questions tend to expose where a company stands. What has become abundant in my business that I still price as if it were scarce? What has become scarce that I am not yet measuring? And of those new scarcities, which do I genuinely control, and which has a platform, a compute provider, or a competitor already captured without my noticing?

The trap is almost always the same. Faced with abundance, the instinct is to fight harder on the abundant layer, to produce more of it, faster. That is exactly the layer where margins collapse toward zero, because everyone there is holding the same tools. The value migrates to the bottleneck and settles with whoever holds it.

The freelancers who recovered did not win by producing more words. The ones now thriving repositioned around what stayed scarce. Those with AI-complementary skills earn around 40% more than peers who don’t, and higher-value contracts have risen. They went back to selling a scarcity.

AI will keep making output abundant. That is its promise, and much of it is progress. But abundance is not where margin lives. It moves to whatever stays scarce: access that’s hard to win, trust that’s hard to earn, power and priority that can’t simply be conjured.

The companies that produce the most with AI will not capture the value. It goes to whoever can name the new scarcity, control it, and price it before a competitor does.

This is a discipline, not an instinct. Every strategy now needs a scarcity map: what the company is making abundant, where that creates a bottleneck, who controls it, and whether the price reflects any of it. The hard part is building the business around that scarcity instead of the output AI has made cheap, while every incentive still pulls toward producing more.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.