AI scarcity shifts to energy, water and attention, commentary argues

A Fortune commentary on AI scarcity argues that generative tools are not removing business constraints so much as moving them. The piece points to water, electricity, compute access, trust and human attention as the places where value may collect as machine-made output becomes cheaper.

The argument starts with data centers. The Lincoln Institute of Land Policy has reported that Texas data centers could draw as much as 399 billion gallons of water a year by 2030, compared with 49 billion gallons in 2025. The Fortune commentary says that kind of demand shows how AI’s low-cost digital output depends on scarce physical inputs, including cooling water in drought-stressed places.

Electricity is another constraint. The International Energy Agency said data-center power use rose 17% in 2025 and is on track to roughly double from about 485 terawatt-hours that year to around 950 terawatt-hours by 2030. The IEA also expects electricity demand from AI-focused data centers to triple, according to the commentary.

What does AI make scarce?

The commentary’s answer is that AI makes routine output more plentiful while increasing the value of resources and assurances that are harder to reproduce. Those include grid connections, priority access to computing capacity, reliable data provenance, regulatory accountability, customer attention and in-person experiences.

Live Nation’s 2025 results are used as one example of demand shifting toward scarce experiences. The company reported a record 159 million fans at its shows and more than $25 billion in revenue, with international attendance exceeding U.S. attendance for the first time.

The labor-market example comes from online freelancing. A study cited by the Complexity Science Hub found that after ChatGPT’s release, demand for tasks well suited to generative AI fell quickly on a global freelancing platform. Translation into Western European languages dropped about 30%, while demand for highly commoditized writing such as “About Us” pages fell by half, according to the study.

Brookings, examining the same market, found that experienced and higher-priced freelancers were hit at least as hard as other workers. The Fortune commentary argues that their work was punished by abundance rather than by a sudden loss of skill: text that once signaled competence became far cheaper to produce.

Who controls the bottleneck?

The commentary says companies using AI can mistake lower production costs for full control of value. If a business depends on outside compute, power, transformers, gas turbines, water or critical minerals, the durable margin may shift to whoever controls those inputs.

The IEA has said data centers in the United States could consume more electricity by the end of the decade than the production of aluminum, steel, cement and chemicals combined. Data Center Dynamics, citing IEA figures, reported that China refines about 99% of gallium, a material used in parts of the technology supply chain.

Water is also becoming part of the same calculation. The MOST Policy Initiative has projected that direct U.S. data-center water consumption could double or more by 2028, and the Fortune commentary notes that new facilities are often being built in basins that already face shortages.

On the demand side, the bottleneck may be attention rather than infrastructure. The commentary cites economist Herbert Simon’s 1971 observation that abundant information creates scarce attention, then applies it to markets where customers rely on platforms, channels or assistants to filter choices.

The strategic lesson, according to the commentary, is that companies should identify what AI has made abundant, what has become scarce, and whether they control that scarce asset. It says pricing should increasingly reflect priority, trust, accountability and access, not only the cost of producing more output.

This story draws on original reporting from Fortune.