The newsroom · Source-backed insight

The Honest Data-Center Debate Is Gross Versus Net

National electricity demand establishes the scale of the compute buildout. It does not settle whether a particular local project is a good deal. That depends on the power, water, ownership, revenue and exit terms put in writing.

By Singularity Farmer · Published September 24, 2026

More compute is coming. Communities should own more of the upside.

That requires separating two questions that are often collapsed into one:

1. Gross: How much infrastructure, electricity, water and land might data centers require?
2. Net: After the enforceable local terms are counted, what does a particular community receive, risk and retain?

This is an editorial framework, not a standardized accounting formula. It prevents national demand figures from being mistaken for project-level outcomes. It also prevents promised benefits from being treated as delivered results.

Start with the national scale

Sourced fact: The International Energy Agency estimates that data centers consumed around 415 terawatt-hours of electricity in 2024, about 1.5% of global electricity consumption. Its Base Case projects approximately 945 TWh in 2030, just under 3% of global electricity consumption. The 2030 figure is a scenario, not a measurement or a forecast for any individual facility. Source: IEA

The United States has its own range. Berkeley Lab estimates that U.S. data centers used 176 TWh in 2023. Its 2024 report models a 2028 range of 325 to 580 TWh. That range describes national scenarios, not the load of a proposed local project. Source: Berkeley Lab

These figures make the infrastructure opportunity legible. They do not answer who pays for a substation, whether households bear rate risk or what happens if a facility closes.

Those answers live in project documents.

Build a local power ledger

Proposal: Local leaders should ask for a project-specific account of the power arrangement before debating broad claims about benefits or burdens.

That review should ask:

• What load is requested, and under which operating schedule?
• Who pays for generation, transmission, substations, storage and backup equipment?
• Which costs can enter the general rate base?
• Who bears curtailment, fuel-price and construction-overrun risk?
• Are any resilience or emergency-service commitments contractual?
• What happens to dedicated infrastructure if the project changes ownership or closes?

The goal is not to make construction impossible. It is to assign costs honestly while expanding useful computing capacity.

Use two water numbers, not one

Water debates become unreliable when withdrawal and consumption are treated as synonyms.

Sourced fact: The U.S. Geological Survey explains that not all withdrawn water is lost because much of it returns to the environment. USGS defines consumptive use as the portion removed from the immediate water environment through processes including evaporation, transpiration, incorporation into products or crops, and human consumption. Source: USGS

Proposal: A local review should request both withdrawal and consumptive-use figures. It should also identify the cooling design, water source, seasonal assumptions and operating conditions behind those numbers.

One unlabeled figure is not enough to evaluate a project.

Count only binding community value

A press release can describe an aspiration. A contract assigns an obligation.

Proposal: Put the proposed local value into a written ledger:

• taxes, abatements and payment schedules;
• ownership or revenue-sharing opportunities;
• local contracting and training commitments;
• funded power or connectivity improvements;
• enforceable limits covering noise, lighting and construction traffic;
• monitoring and public reporting requirements; and
• financial responsibility for modification, decommissioning and site restoration.

Each item should have a responsible party, a deadline, a measurement method and a remedy for nonperformance. Communities should not count projected jobs, lower rates, water savings or resilience benefits as established outcomes without the applicable agreements and engineering record.

Negotiate the net

Editorial thesis: Data centers can be good infrastructure when communities capture durable value from them. More capacity is the goal. A better deal is the method.

The gross national numbers explain why the buildout deserves attention. The net local terms determine whether a particular proposal deserves approval, revision or rejection.

Welcome the infrastructure. Inspect the design. Put the community’s terms in writing.