Back to the episode map

Evergreen

Data Center Community Impact: Records to Request

Communities should evaluate a data center with project-specific power, water, cooling, air, land, incentive, workforce, emergency, and reporting records.

Aug 4, 20266 min readBy Dalton Anderson

What Communities Should Ask Before Approving a Data Center

A community cannot evaluate a data center from a national electricity average, a promised job count, or a company sustainability report. It needs project-specific records for power, water, cooling, backup generation, emissions, land, tax incentives, construction, workforce, emergency response, and continuing public reporting.

The central question is not whether data centers are good or bad. It is whether the proposed facility's costs, benefits, dependencies, and protections are visible enough for a public decision.

National context is not a site estimate

The Department of Energy's 2024 data-center electricity report summarized Lawrence Berkeley National Laboratory scenarios for rapidly rising U.S. demand. DOE's current data-center resource hub collects later federal material.

Those sources describe a national system. They cannot establish one project's peak load, annual use, ramp schedule, grid connection, generation source, or effect on local rates.

The Government Accountability Office's 2025 assessment says estimates of generative AI's environmental effects vary because companies disclose limited information about infrastructure, energy, emissions, and water. That uncertainty is a reason to require project records, not a reason to substitute a dramatic national number.

Begin with the actual facility

The application should identify the legal applicant, property owner, operator, end user if known, site, acreage, buildings, phases, planned IT load, total facility load, operating schedule, cooling design, expected useful life, and decommissioning responsibility.

A campus announced as 100 megawatts may have a different total load once cooling, power conversion, lighting, and support systems are included. A later phase can change the impact after the first approval.

flowchart TD
    A["Proposed data center"] --> B["Power and grid"]
    A --> C["Water and cooling"]
    A --> D["Air and backup generation"]
    A --> E["Land, noise, traffic, and emergency response"]
    A --> F["Taxes, incentives, jobs, and public cost"]
    B --> G["Project-level public record"]
    C --> G
    D --> G
    E --> G
    F --> G
    G --> H["Approval terms, monitoring, and enforcement"]

Request the power record

The record should include expected peak megawatts, annual megawatt-hours, phase timing, load factor, interconnection study, utility service agreement, transmission and distribution upgrades, generation assumptions, demand-response obligations, outage priorities, and responsibility for capital and ongoing cost.

The community should be able to distinguish company-funded infrastructure from costs recovered through other customers. Confidential commercial details may need protection, but a confidentiality claim should not erase the material public allocation.

If dedicated generation is proposed, request the fuel, capacity, operating profile, construction permit, emissions controls, ownership, and retirement or conversion plan.

Request the water and cooling record

Water use depends on cooling technology, weather, workload, water quality, reuse, and operating choices. The application should state source, withdrawal authorization, peak daily and annual use, consumptive use, discharge, treatment, chemicals, sewer capacity, drought operation, and alternatives.

The record should separate potable, reclaimed, surface, and groundwater. It should show how the project performs during drought or extreme heat, when cooling demand and community water stress may rise together.

Water withdrawal, consumption, and discharge are different quantities. A large withdrawal can return much of the water. Evaporative cooling can consume water by returning it to the atmosphere. Neither should be described with a single unexplained gallons figure.

Request air and backup-power records

Backup generators can create a material local air source when installed in large numbers or tested regularly. Request the generator count, manufacturer, model, capacity, fuel, stack, emission factors, expected testing hours, emergency-use definition, permits, controls, monitoring, maintenance, and cumulative site emissions.

The Environmental Protection Agency's stationary engine compliance resources explain federal requirements for certain engines. Federal rules do not replace state and local permits.

Do not infer health outcomes from the equipment list alone. A site-specific review needs emissions, dispersion, background conditions, exposure, population, operating history, and qualified public-health evidence.

Price the public exchange

Economic-development claims should identify permanent jobs, construction jobs, wages, occupations, residency assumptions, vendor spending, tax revenue, exemptions, abatements, grants, land discounts, infrastructure subsidies, and public-service cost.

A job promise should have a date, definition, reporting method, and remedy if it is missed. Construction employment should not be presented as permanent operating employment.

Tax agreements should show the baseline, term, renewal, transfer, clawback, audit rights, and the public body responsible for enforcement.

Plan for ordinary and extreme operations

Request the fire and emergency plan, hazardous materials record, battery system design, responder access, water supply, mutual-aid requirements, evacuation or shelter assumptions, cybersecurity coordination, severe-weather plan, outage behavior, and incident notification.

FEMA's community lifelines framework can help a locality consider how energy, communications, water, transportation, safety, and other services interact during disruption. It does not replace a facility-specific plan.

Noise, lighting, traffic, construction hours, road wear, stormwater, wetlands, habitat, and cultural resources also need the records required by the site and jurisdiction.

Make approval measurable

An approval should identify which statements are enforceable, who monitors them, which data is reported, how often, where the public can see it, what counts as a violation, and which remedy applies.

Useful continuing records can include monthly electricity and water use, cooling source, generator operation, reportable incidents, employment, incentive compliance, complaints, permit exceedances, and expansion requests.

The data should be specific enough to compare the operating facility with the approved case while protecting legitimate security and customer information.

The record should be public. It should be easy to find. It should use the same units each month. Each phase should have its own line. Changes should be dated. Missing data should be marked. Claims should link to the permit or agreement that controls them.

Avoid the national-average trap

One facility can use less water because it relies on air cooling and more electricity as a consequence. Another can use reclaimed water but place new demand on treatment infrastructure. A project can create substantial tax revenue while shifting grid costs. A low permanent job count may still accompany a strategic infrastructure benefit.

Those are tradeoffs to measure, not slogans to resolve.

Episode 102 raised serious concerns about undisclosed water use and backup turbines. The public guide does not repeat specific allegations from the recording without permits, utility records, enforcement records, and qualified evidence for the named site.

The decision standard

Ask for the design basis, not only the headline. Ask who pays, who benefits, who monitors, and what happens when the project changes.

A community can support, oppose, modify, or condition a proposal. Whatever the decision, it should rest on a record that can still be audited after the ribbon cutting.

AI assisted with research organization, structure, drafting, and validation. Dalton Anderson remains the attributed author and final editorial authority. The transcript and linked public sources control factual claims. Publication remains unauthorized.

Sources

Follow the evidence.

  1. epa.gov: stationary enginesepa.gov
  2. gao.gov: gao 25 107172gao.gov
  3. spec.c2pa.org: Explainerspec.c2pa.org
  4. cdc.gov: fatigue workcdc.gov
  5. Current Google Search AI feature documentationdevelopers.google.com
  6. pubmed.ncbi.nlm.nih.gov: 22708885pubmed.ncbi.nlm.nih.gov
  7. fema.gov: lifelinesfema.gov
  8. pubmed.ncbi.nlm.nih.gov: 31864153pubmed.ncbi.nlm.nih.gov
  9. Google's guide to optimizing for generative AI featuresdevelopers.google.com
  10. energy.gov: powering americas ai future data center resource hubenergy.gov
  11. energy.gov: doe releases new report evaluating increase electricity demand data centersenergy.gov
  12. deepmind.google: synthiddeepmind.google
  13. content.naic.org: naic adopts first national climate resilience strategy insurance close coverage gaps and improvecontent.naic.org
  14. steveblank.com: whats a startup first principlessteveblank.com
  15. content.naic.org: resilience data visualization mapcontent.naic.org
  16. developers.google.com: google search and ai contentdevelopers.google.com
  17. ycombinator.com: jared friedmanycombinator.com
  18. Introducing Search Generative AI performance reportsdevelopers.google.com
  19. nist.gov: reducing risks posed synthetic content overview technical approaches digital contentnist.gov
  20. ycombinator.com: startup school videosycombinator.com
  21. spec.c2pa.org: charterspec.c2pa.org
  22. swissre.com: Natcat protection gapswissre.com
  23. fema.gov: requirementsfema.gov
  24. harpercollins.com: the hard thing about hard things ben horowitzharpercollins.com
  25. ibhs.org: wildland fire embers and flames home mitigations that matteribhs.org
  26. dash.harvard.edu: 13a7b031 0fdd 45ec a7e0 2b80e2bc679fdash.harvard.edu
  27. ycombinator.com: 8g how to get startup ideasycombinator.com
  28. deepmind.google: identifying ai generated images with synthiddeepmind.google
  29. pubmed.ncbi.nlm.nih.gov: 33749936pubmed.ncbi.nlm.nih.gov
  30. csrc.nist.gov: finalcsrc.nist.gov
  31. nibs.org: NIBS MMC MitigationSaves 2019nibs.org
Data Center Community Impact: Records to Request