Data centers and the community balance sheet: a transparent project-rating framework

The question this investigation answers

Data centers can create construction work, strengthen a local tax base, support useful digital services, and attract investment. They can also require major electricity infrastructure, alter land and water systems, and leave a community carrying costs that are difficult to see in a groundbreaking announcement.

A serious assessment has to measure both possibilities at the same time. The question is not whether computing is valuable in the abstract.

The question is what a particular project, in a particular place, will deliver, consume, cost, and contractually guarantee.

This framework is CrownThrive's proposed editorial assessment method. Its weights and publication thresholds are original policy choices, open to public challenge and versioned revision.

They are not government certifications, engineering approvals, investment recommendations, or claims that any named project has passed. No project receives a rating until its actual evidence has been collected.

The purpose is to make a developer's strongest case, a resident's strongest concern, and an institution's formal findings comparable without flattening them into a popularity contest.

Why the assessment must follow the whole project

A data center is more than a building full of servers. Its practical footprint includes substations, transmission connections, water supply, backup power, roads, construction staging, telecommunications, contractors, financing, and eventual equipment replacement.

A campus built in phases can change its impact substantially after the first zoning hearing. Evaluating only the first building can therefore miss the infrastructure required for the completed site.

Historical national numbers establish scale but cannot substitute for local measurements. The Department of Energy's December 2024 summary of Berkeley Lab research estimated U.S. data-center electricity consumption at 176 terawatt-hours in 2023, or about 4.4 percent of national consumption.

Its 2028 figures were projections, not measurements of completed development. Those figures remain dated reference points; this scorecard does not present them as current 2026 consumption.

(Department of Energy, 2024 report summary)

Virginia's legislative audit offers a useful regional warning against simplistic accounting. Its 2024 study found substantial economic benefits, concentrated heavily in construction, and examined the infrastructure challenges created by rising electricity demand.

It also distinguished current cost allocation from future system-cost pressures. A project can pay its assigned tariff today while growth still changes costs across the wider system.

These are different questions that require different evidence. (Virginia JLARC, Data Centers in Virginia)

The unit of assessment is therefore a named project phase, tied to an identifiable site, developer, operator where known, utility service territory, watershed, permit record, and reporting date. A planned facility and an operating facility use the same categories, but their status remains visible.

A promise receives a promise label. A completed upgrade receives a completion record.

A measured outcome receives a measurement period. Changing the status of a record never silently erases its earlier version.

Four dimensions and a public weighting rule

The default score is a 100-point community-impact assessment. It assigns the greatest weight to infrastructure and grid impacts because those obligations can extend well beyond a property boundary.

Water and land receive a separate substantial weight because an annual utility bill does not capture watershed stress, agricultural conversion, or construction runoff. Clean energy and community benefits receive their own dimension, and workforce development and local sourcing are assessed independently of capital investment headlines.

Readers should be able to change these weights in a scenario view, but the published default rating must retain the original weights. A developer may prefer a heavier emphasis on tax receipts; residents near a stressed water source may emphasize water.

Both scenarios are informative. Neither may quietly replace the common comparison.

Every displayed variant must identify its weights and be labeled a reader scenario rather than the official editorial score.

Each criterion is rated from zero to five using published anchors. Zero requires evidence of a substantial failure or absence of a required control.

One indicates weak protection with major documented gaps. Two means a partial response.

Three indicates a credible, documented approach that meets the stated assessment standard. Four requires stronger performance with measurable commitments and a verification path.

Five requires demonstrated, sustained performance against demanding targets and independently checkable results. A missing document is marked unknown, not automatically zero and never automatically acceptable.

Before assigning a criterion score, the evaluator must publish its scoring sheet: the local baseline, target, unit, measurement period, evidence requirements, and reasons for the selected thresholds. There is no universal water-volume threshold that makes sense for every watershed, and no universal employment number that establishes a good local outcome.

Changing a threshold after seeing a favored project's result requires a new methodology version and a recalculation of comparable projects. Planning-stage assessments must be labeled prospective and compared with other planning-stage assessments; an unbuilt commitment cannot receive a top outcome score reserved for demonstrated operation.

• Infrastructure and local grid impact — 30 points: Credible capacity planning, clear cost responsibility, tested flexibility, efficient operation, and neighborhood impact controls
• Water and land stewardship — 25 points: Measured water demand, drought resilience, appropriate siting, and verified runoff and ecological controls
• Clean energy and community benefits — 25 points: Traceable energy claims, additional local benefits, transparent fiscal terms, and enforceable delivery
• Workforce development and local sourcing — 20 points: Durable employment, quality training outcomes, accessible procurement, and documented local value retained

Infrastructure: capacity, cost, flexibility, efficiency, and neighbors

The grid dimension is divided into five criteria: connection and capacity planning, eight points; cost allocation and exit protection, eight; load flexibility and resilience, six; facility efficiency, four; and local infrastructure and neighborhood controls, four.

Start with the load boundary. The record should distinguish contracted power, maximum requested load, installed IT capacity, measured peak demand, average demand, and electricity consumed over a stated period.

These terms must not be combined into a single megawatt figure. Megawatts describe power at a point in time; megawatt-hours describe energy across time.

A very large requested connection is not evidence that the entire amount is already being consumed. Conversely, a low initial load does not describe the impact of a fully developed campus.

A transparent connection record identifies the project phase, expected energization date, responsible utility, relevant study or docket, required substation and transmission work, and any conditions that must be satisfied before demand increases. Public reporting should use summaries sufficient for accountability without publishing sensitive diagrams, protective settings, or details that could facilitate an attack.

The score concerns whether the plan has been demonstrated and who is accountable for it, not whether the public can see every engineering drawing.

Cost allocation asks who pays if the project arrives, expands, delays, or leaves. Record the contribution to construction, tariff structure, minimum payment obligations, security or collateral where disclosed, recovery period, and responsibility for stranded infrastructure.

Separate the utility's approved allocation from an editorial estimate of distributional effects. A forecast of future residential bills must identify its model, assumptions, geography, uncertainty, and publication date.

It cannot be inferred from a company's capital-spending announcement alone.

Load flexibility should be described as an operational capability with terms: megawatts available, notice required, duration, recovery period, annual call limit, contractual availability, and evidence of a completed test. A campus may be able to pause some batch computing while maintaining time-sensitive services.

The assessors should not assume that all AI work can move freely across time or geography. Workload constraints, connectivity, service commitments, and data restrictions can matter.

Credit is earned for verified capability, not for a generic statement that software is flexible.

Efficiency requires a clearly defined measurement boundary. The familiar facility-to-IT electricity ratio is useful only if numerator and denominator cover the same time and facility.

A design estimate should not be compared directly with a year of measured operation. Publish total electricity as well as the ratio: an improving ratio can coexist with rising total consumption.

Weather, utilization, commissioning, and equipment mix can affect comparisons. The score should reward transparent improvement while preserving the community's ability to understand the absolute load.

The neighborhood assessment follows noise, traffic, road wear, construction hours, backup-generator operation, and emergency coordination. For noise, retain measurement location, time, operating condition, instrument method, and the applicable local standard.

A resident report is valuable evidence of an experience, but it is not interchangeable with a calibrated sound survey. For backup power, distinguish permitted capacity, test runs, emergency operation, fuel type, and measured or reported emissions.

An installed generator is not evidence of continuous operation.

EIA offers public electricity series, including balancing-authority operating data and retail price information. These are useful for regional context, but they do not reveal a private facility's individual meter readings.

The data pipeline must preserve that geographic distinction instead of turning a regional trend into a project-specific claim. (EIA Open Data)

Water and land: measure the place, season, and full boundary

Water and land use receive five criteria: demand and disclosure, seven points; drought resilience and source protection, six; land conversion and siting, five; stormwater and ecological protection, four; and cumulative impacts, three.

The water ledger needs withdrawals, consumption, discharge where relevant, source type, reporting period, and operating conditions. Withdrawal means water taken from a source; consumption concerns the portion not returned within the relevant accounting boundary.

Those concepts answer different questions. A cooling system that circulates water internally may still require replacement water.

A claim of closed-loop operation therefore needs a boundary and an explanation, rather than an assumption that the entire site uses no water.

Publish annual volume alongside the hottest-month demand and the maximum-day design requirement. A community may have adequate annual supply but face pressure during a drought or heat wave.

The scorecard should ask what triggers a reduction, who can require it, whether an alternate cooling mode exists, what that mode costs in electricity, and whether it has been tested. An alternative source must be available under the same stressed conditions; a plan that depends on an unapproved future pipeline remains a future plan.

Water intensity should be shown with its denominator. Liters per IT kilowatt-hour, liters per facility kilowatt-hour, and liters per computational task are not equivalent.

The assessment should retain the original figure and normalize only when the underlying inputs are available. Indirect water associated with electricity generation belongs in a separate estimate with a disclosed method.

It should not be blended with directly metered water into a single apparently precise number without explaining the assumptions.

Watershed context can be informed by public monitoring. USGS Water Data for the Nation provides monitoring-location records and continuous or other observed water data.

A nearby gauge may help explain conditions, but its relationship to the facility's actual water source must be established. Proximity alone is insufficient proof that two users draw from the same constrained supply.

(USGS Water Data for the Nation)

Land analysis begins with acreage and previous use. Record the site boundary, disturbed area, new impervious surface, prior agricultural or industrial use, forest removal, habitat information, and relevant flood or environmental reviews.

Reusing an industrial parcel can reduce some kinds of new disturbance while introducing remediation obligations. A rural greenfield may offer space while changing agricultural access or drainage.

The score should reveal these differences instead of awarding points simply because a project is urban or rural.

Stormwater is a construction and operating issue. The evidence file should contain the applicable permit, responsible operator, inspection records where public, control plan, and documented correction of problems.

EPA explains that construction runoff can carry sediment, debris, and chemicals into water systems. The framework uses that physical pathway to require project-specific inspection evidence; it does not assume that every permitted site is causing contamination.

(EPA, Stormwater Discharges from Construction Activities)

Cumulative assessment asks what happens when neighboring projects reach the same stage together. Report the other projects included, their status, and the scenario date.

Do not add speculative proposals to operating demand as if all were certain. Present separate operating, committed, and proposed totals.

A sensitivity analysis should show whether the conclusion changes if a major project is delayed or never built.

Clean energy and community benefits: follow the contract to the outcome

This dimension includes energy and emissions disclosure, six points; additional clean-energy delivery, seven; enforceable community benefits, seven; and transparent fiscal value, five.

The energy ledger separates electricity used at the site, contractual purchases, certificate claims, generation owned or financed, storage, backup generation, and any exported energy. An annual contractual matching claim does not, by itself, describe the hourly physical supply serving a facility.

Both views may be useful, but they should remain visibly distinct. A project should identify the accounting method, geographic relationship, operating date of associated generation, and how attributes are tracked to prevent multiple claims over the same benefit.

Emissions estimates should state whether they are based on facility measurements, fuel records, supplier factors, or regional averages. EPA's eGRID contains electricity generation and emissions information that can support a regional estimate.

Such an estimate is not an hourly dispatch model and does not establish the exact emissions caused by a new individual project. Use the appropriate factor, identify its vintage, and preserve the calculation.

(EPA eGRID)

Clean-energy credit should recognize delivery risk. An announced power purchase, an executed contract, a financed project, a commissioned generator, and verified energy delivery represent different stages.

A future plant should never be counted as current production. Storage should disclose both power capability and energy duration; a battery's megawatt rating alone does not say how long it can support a load.

Proposed technologies need a commissioning timeline, performance basis, and contingency plan rather than a presumption of commercial readiness.

Community benefits need named recipients, deliverables, dates, funding sources, and enforcement. A scholarship fund can be useful; so can a road improvement, training partnership, public connectivity project, or emergency-service contribution.

The score asks whether the benefit is additional, accessible, and delivered. A mandatory payment should not be described as a voluntary gift.

A benefit funded through a public subsidy should disclose that source. The same payment cannot appear in multiple categories as separate value.

Fiscal analysis should show gross local receipts, incentives and abatements, public expenditures, the measurement period, and uncertainty. Capital investment is not local tax revenue.

A taxable equipment base changes over time, and expected receipts may depend on replacement cycles and policy choices. The public record should include a plain-language explanation of those assumptions, along with a downside scenario.

Communities need to know how dependent a budget could become on one operator or a narrow set of assets.

Workforce and sourcing: count careers, completion, and retained value

Workforce and local sourcing include job quantity and quality, seven points; training and advancement, five; local procurement, five; and access and accountability, three.

The jobs ledger must separate construction job-years, peak people on site, permanent full-time equivalent positions, contractors, and indirect estimates. Ten people employed for one year and one person employed for ten years can produce the same job-year total while creating very different local experiences.

Both duration and timing belong in the record. Publish wage distributions or bands, benefits, working schedules, turnover, and local residency definitions where reliable data is available.

Training should be scored on more than enrollment. Record paid training time, completion, credentials earned, placement, wage progression, and retention at agreed intervals.

An employer's scholarship commitment is distinct from a graduate obtaining a relevant job. Registered apprenticeship provides an established model combining paid work and structured learning; a local program should be described accurately rather than using the term for an unverified short course.

(Apprenticeship.gov, Registered Apprenticeship Program)

Local procurement requires a fixed geography and a defensible denominator. Measure eligible spending placed with qualifying local businesses, disclose excluded categories, and identify whether a vendor is actually performing work locally.

A mailing address is not sufficient proof of local economic activity. Materials sourcing should distinguish extraction, manufacture, distribution, installation, and professional services.

Buying imported equipment from a local reseller creates some local value, but not the same local production value as manufacturing it there.

Small suppliers need a practical route into the opportunity. The assessment records published opportunities, bid requirements, payment terms, insurance demands, subcontracting access, and a route for resolving late payment.

A high local-spend percentage can conceal exclusion if only one large incumbent can qualify. Accessible procurement is a separate outcome from gross expenditure.

A community should be able to trace whether an announced supplier initiative created completed contracts and timely payments.

Evidence, uncertainty, and the public score

Evidence quality is reported beside impact performance rather than quietly multiplied into it. Level A is independently verified project evidence with an accessible method and scope.

Level B is an accountable primary record, such as a permit, utility filing, or measurement series, appropriate to the claim. Level C is an attributable operator document or commitment without independent outcome verification.

Level D is an attributed stakeholder report requiring corroboration. Level U means missing, unusable, or undisclosed evidence.

An official filing can still contain a forecast, and an audit can still have a limited scope; the label never replaces reading the record.

The numerical rule is deliberately simple. For criterion weight w and performance score s from zero to five, the contribution is w × s ÷ 5.

Weights sum to 100. For unknown criteria, publish a range: the lower bound includes established contributions, while the upper bound adds all points still unresolved.

These bounds express uncertainty, not a claim that unknown performance is zero or excellent. Display weighted coverage separately as the share of criteria supported well enough to assess.

A single headline rating is permitted only when at least 80 percent of weighted criteria are assessable, every dimension has at least 60 percent coverage, and critical grid, water, permitting, and cost-responsibility questions have sufficient evidence. Otherwise the page says “Insufficient evidence for a headline rating” and publishes the available facts and score interval.

These thresholds are this publication's methodology choices. They are not statutory approval rules.

Even a numerically strong project can carry a critical flag. A verified unresolved safety breach, material misrepresentation, required permit gap, or severe documented impact remains prominent and cannot be canceled by unrelated benefits.

Allegations are labeled allegations; findings are linked to the responsible authority or evidence. Responses and corrections remain visible.

Advertising, sponsorship, or payment for a directory listing never changes the score, the evidence standard, or a project's right to reply.

How the tracker stays useful after publication

Each project page should show its phase, last review date, next review trigger, source register, open questions, and change history. Metered information follows the meter's reporting cycle.

Permit and hearing records follow their official publication. Annual jobs and procurement commitments receive annual outcome checks, with intermediate updates when credible evidence becomes available.

A feed failure produces a visible stale-data state rather than silently refreshing the timestamp on old information.

Residents, developers, utilities, workers, farmers, conservation groups, local businesses, and public officials should all have a route to submit evidence and respond. The same factual claim receives the same scrutiny regardless of who submits it.

Public polls can document participating readers' views, but a self-selected poll is not proof of the opinion of every resident. It also cannot establish whether an engineering design works.

The community record becomes stronger when lived experience, measured outcomes, and institutional findings are all present with their differences intact.

The result is a practical public ledger: what is proposed, what is promised, what is permitted, what has been built, what has been measured, and what remains unresolved. That sequence gives a community a fair way to recognize good development, demand correction where performance falls short, and revisit conclusions when the evidence changes.