Scarcity’s Hidden Patterns

The widespread assumption that artificial intelligence (AI) has abolished scarcity is misleading. What we’ve actually seen is the relocation of scarcity from one place to another.

Consider Texas, where droughts are a recurring emergency. Data centers driven by AI’s insatiable need for computational power and cooling are projected to consume an astonishing 399 billion gallons of water per year by 2030. This translates to enough water to lower Lake Mead, the largest reservoir in the country, by over sixteen feet annually.

Meanwhile, as AI-generated content has become ubiquitous, demand for human experiences – live events and concerts – has surged. In 2025, a record-breaking 159 million fans attended Live Nation shows, generating over $25 billion in revenue. For the first time, more of these attendees came from outside the United States than within it.

These two phenomena are connected by scarcity. The abundance we see on our screens and devices is not matched by an equivalent abundance of human attention or physical resources. AI has merely relocated the constraints from digital output to the underlying inputs it consumes – water, electricity, and other physical resources.

A recent study on freelancing platforms highlights this dynamic. As AI began to offer cheap, high-quality content, demand for tasks that require human judgment and expertise plummeted. The most experienced and highest-priced freelancers were hit particularly hard, their skills no longer seen as valuable in a world where output had become abundant.

Abundance can be a smokescreen for scarcity. While it’s true that AI has made certain tasks and outputs more accessible than ever before, this has come at a cost. The real constraint is no longer the production of competent text but rather the judgment and accountability that accompany it. What was once scarce – human expertise and attention – now becomes even scarcer as a result of AI’s abundance.

The relocation of scarcity has far-reaching implications for businesses and organizations alike. In the industrial economy, scarcity was often visible and easy to quantify: inventory levels, engineer-hours, or seats available. But in the AI-driven world, these constraints become diffuse and sometimes invisible.

For instance, a streaming service may be able to surface a thousand titles in an instant but is actually rationing the viewer’s time and attention. A logistics platform may price every route in milliseconds only to run into capacity limits that have nothing to do with computing power: the number of trucks idling at a single loading dock.

Herbert Simon’s insight from 1971 remains as relevant today: an abundance of information produces a scarcity of attention. As entire sectors operate under this new paradigm, we see the effects – congestion, frustrated customers, trade-offs nobody ever chose out loud. But many organizations still reason as if scarcity had been abolished, ignoring the very real constraints they face.

The question then becomes: who owns the new scarcity? In a world where companies drive their marginal costs toward zero, they often merely push the bottleneck up the chain – to whoever controls the compute and power it runs on. The capital spending of major tech firms has skyrocketed in recent years, with data centers consuming increasing amounts of electricity.

The constraints now bind at the grid level, with direct consumption by U.S. data centers projected to double or more by 2028. This isn’t an exception – reservoirs draining in Texas are merely a harbinger of what’s to come. As we look toward the end of this decade, it becomes clear that AI’s growth has created new scarcity patterns that can no longer be ignored.

“There is, as they say, no AI without energy,” says the IEA’s director. This means companies congratulating themselves on near-free production have, in many cases, merely handed the durable margin to whoever owns the scarce input – water, electricity, or other physical resources. As we move forward, we must acknowledge and address these new scarcity patterns rather than perpetuating a myth of abundance. Only then can we begin to grasp what it means to live and operate in an AI-driven world.