The steelman for data centers: what the industry gets right

This is the piece to read if you suspect Stand is a one-sided site. It is not. The data-center industry — through trade groups like the Data Center Coalition and through individual operators' public filings — makes several claims that hold up under scrutiny. Residents who skip past the strongest version of the industry's case argue from a weaker position. Here is the strongest version, plainly stated, followed by where resident concerns remain valid even under it.

Why steelmanning matters

A steelman is the opposite of a strawman. Instead of attacking the weakest version of the other side's argument, you state their strongest version — and then engage with that. Courts, commissioners, and neighbors are more persuaded by someone who clearly understands the counterargument than by someone who dismisses it. If Stand's only goal were to stop every data center in Texas, a one-sided article would do the job. Stand's goal is better decisions, not reflexive ones. So here is the case, as industry would make it.

Claim 1: AI and cloud demand are real and growing.

This is verifiable from independent sources. Global data-center electricity consumption rose from roughly 1 percent of global demand in 2018 to an estimated 2 percent in 2024, and the International Energy Agency projects it will roughly double again by 2030, driven heavily by AI training and inference workloads. In the United States, data-center electricity demand is projected to reach 8 to 12 percent of total U.S. generation by 2030, up from about 4 percent today.

If the workload exists, physical facilities have to house it. The question "should this data center be built anywhere?" is not really open. The question "should this one be built here, with these conditions?" is — and that is the question a commissioners' court actually decides.

Claim 2: Closed-loop cooling is a real option, and the best operators use it.

Not every data center is worst-case on water. Meta has publicly committed to closed-loop cooling for new builds in water-stressed regions, and Google's 2024 Environmental Report documents facilities operating at near-zero evaporative water use through closed-loop and air-cooled designs. Some hyperscale campuses in Arizona and Nevada — sited in literal deserts — have been specifically engineered to avoid evaporative cooling entirely.

Residents who argue "all data centers use millions of gallons" lose credibility at the microphone when the specific project in front of them has disclosed air-cooled or closed-loop cooling. The industry is right that the range is wide. The seven-question checklist works here: ask for the number, and if the answer is small, engage accordingly.

Claim 3: Permanent jobs are modest, but construction is real.

A typical hyperscale campus employs 30 to 150 permanent staff at full operation — a real but small number for a facility of that size and cost. The industry does not usually hide this. Where industry's case is stronger is on the construction phase: a single campus generates hundreds of construction jobs at peak, often 400 to 1,200 workers, for 18 to 36 months. Those wages are real income for local workers, and construction sales tax is real revenue for the local and state budget.

The honest question is whether the lifecycle value — construction phase plus ten or twenty years of partial property-tax payments under an abatement, plus the eventual full-tax years — clears the cost of the infrastructure, abatement, and externalities. That is a fiscal-modeling question a competent county can answer. It is not a question that should be waved off in either direction.

Claim 4: Some jurisdictions see real net fiscal benefits.

Prince William County, Virginia collected more than $1 billion in data-center-related tax revenue in fiscal year 2024 — roughly one-third of the county's total general-fund revenue. Loudoun County, Virginia, home to "Data Center Alley," has used data-center property-tax revenue to fund schools, public safety, and infrastructure at a scale no other industry in the county approaches.

The honest caveat: those outcomes materialized after two decades of concentrated investment, hard-won regulatory expertise, and repeated legal fights. A Texas county approving its first hyperscale project today is not in the same institutional position as Loudoun County in 2024. The outcome range is wide. Some Virginia counties are net winners. Some have sued to get out of the commitments they made. Both data points are real.

Claim 5: Broadband, latency, and AI supply-chain security have real local upsides.

Hyperscale campuses drive investment in regional fiber backbones. Those backbones then serve nearby homes, schools, and small businesses. It is not automatic — the fiber sometimes bypasses the rural communities it crosses — but it is often bundled with the project.

On latency: AI inference workloads for end-users benefit from physical proximity. If the AI tools your hospital, school district, or small business uses are served from a data center 500 miles away, response times are measurably worse than if they are served from 50 miles away. The case for geographic distribution of compute is real, and Texas's power, land, and tax environment make it a reasonable location.

On AI supply-chain security: there is a federal-level argument that hosting AI workloads on U.S. soil, in jurisdictions with robust physical and cyber security, is preferable to hosting them abroad. That argument is real at the national level even if it does not obligate any specific Texas county to say yes to any specific project.

Where resident concerns remain valid under the best industry case

A serious pro-industry case still leaves room for serious concerns. Four of them.

1. The cost-allocation question is genuinely unsettled.

Even industry filings at the PUCT acknowledge that who pays for transmission upgrades — the operator, all ratepayers, or some split — is a policy choice, not a technical given. SB 6 put that policy choice on the PUCT's docket precisely because the Legislature did not decide it. Residents who comment on that docket are not being anti-industry; they are participating in a fight the state explicitly invited.

2. Decommissioning is underaddressed almost everywhere.

No data-center filing the authors have reviewed includes a substantive decommissioning bond. If a facility becomes uneconomic in 20 years — because AI architectures have changed, because power prices have shifted, because the operator defaults — the county is left with the demolition bill. This is a reasonable condition to request, and industry has not yet offered it at scale.

3. Groundwater draws on stressed aquifers remain a problem regardless of cooling efficiency.

A 0.1-MGD closed-loop draw from an already-declining aquifer is still a draw from an already-declining aquifer. Efficient cooling helps. It does not solve the underlying aquifer-depletion math in parts of Texas where wells have already dropped 20 or 30 feet over a decade. Cooling-method disclosure answers one question but not the hydrological one underneath it.

4. Local governance capacity varies dramatically.

A rural Texas county with no planning staff approving its first $1 billion industrial project is in a different institutional position than Loudoun County, Virginia, whose 20 years of experience includes hired expertise, established case law, and institutional memory. The best industry case assumes competent local counter-parties. Many Texas jurisdictions are still building that capacity.

What this means for how to show up

If you read this piece and find yourself agreeing with industry on a specific project, that is a legitimate position. The seven-question checklist still works. A project that answers all seven well should be easier to support. A project that cannot answer them is worth questioning regardless of your starting view.

Stand's bet is that informed residents — on all sides of the specific vote — produce better decisions than either uniform opposition or uniform approval. The seven questions, the public-comment formula, and the procedural knowledge in this Learn hub are built to work whether you show up to support, oppose, or condition a project.