
Data Centers, Energy Use, and Local Backlash
Water, land and local impacts
Environmental impacts: define the system before judging it
Boundary first · Site ≠ supply chain · Analysis framework
A campus is not the whole footprint
Environmental assessment should distinguish the occupied building, the full planned campus, supporting infrastructure, and upstream supply chains. A server room can have little direct combustion while depending on electricity generation elsewhere. Conversely, a facility with on-site generation may shift impacts closer to neighbors without eliminating upstream fuel production. Minnesota identifies environmental review, air permitting, water discharges, stormwater, and fuel storage as distinct oversight areas rather than one interchangeable approval. [51]
Recommended assessment boundary
For this report, evaluate four layers: normal operations; abnormal conditions and backup operations; construction and expansion; and purchased electricity, equipment, and materials. Record which layers each published statistic covers. Treat a corporate annual disclosure as evidence about its stated portfolio, not automatically about an individual campus. Likewise, an approved design describes permitted or proposed activity, not necessarily measured operation.
Compare like with like
A useful impact register names the physical quantity, unit, geography, time interval, and evidence type. Water delivered is not necessarily water consumed. Generator capacity is not annual generator energy. A land purchase is not the disturbed area. A commissioning announcement is not proof that every planned building has reached its advertised electrical load.
Decision implication
The recommended decision sequence is to identify a plausible impact pathway, request project-specific evidence, test mitigation, and assign monitoring responsibility. Neither the absence of an impact estimate nor a broad sustainability commitment should be scored as zero impact. Unknowns belong in a visible evidence queue with an owner and deadline. This makes the analysis useful for supporters, critics, utilities, and local officials without predetermining whether a particular project should proceed.
Water accounting: withdrawal, consumption, and return flows
Separate flows · Use a water balance · Do not compare unlike totals
Three different questions
Withdrawal describes water taken from a source. Consumption concerns the portion not readily available for another local use, including evaporation. Return flow is water returned after use, potentially at another location, temperature, or quality. USGS cautions that organizations sometimes use water terminology differently and that some datasets use withdrawal and water use interchangeably. A label reading simply “water use” therefore needs its underlying definition. [52]
Recommended campus ledger
Request monthly source withdrawals or utility deliveries, wastewater discharge, estimated evaporation, other losses, and storage changes. Identify whether domestic plumbing, landscaping, construction, fire testing, and humidification are included. Metered purchases alone cannot establish consumption unless the return and storage terms are reconciled. For multiple supplies, keep potable, reclaimed, groundwater, and surface water separate before calculating the total.
Illustrative arithmetic, not a project estimate
Suppose a campus receives 100 units of water, returns 30 to the same local system, and has no net storage change. A simplified balance suggests 70 units are consumed or otherwise unaccounted for. That result needs investigation if leaks, transfers, or measurement errors are material. It does not justify claiming all 100 units disappeared.
What should be verified
The analysis should specify where returned water goes and whether it remains usable when other users need it. A return to a downstream river is not necessarily an immediate replacement for groundwater withdrawn from another basin. Ask for meter coverage, calibration dates, estimated fractions, and uncertainty. Report withdrawals and consumption side by side rather than choosing whichever number makes the project look better or worse. Preserve the original source terminology in the evidence record.
Cooling architecture: follow heat from chip to atmosphere
Liquid ≠ evaporative · Trace every loop · Compare peak conditions
Two separate cooling decisions
Removing heat from chips is different from releasing heat outdoors. Direct liquid cooling can move heat from processors through a closed internal circuit, while the facility rejects that heat through dry equipment, an evaporative tower, or a hybrid arrangement. DOE guidance describes both direct liquid systems and heat rejection equipment; the phrase liquid-cooled alone cannot establish water consumption. [53]
Recommended engineering evidence
Request a simplified process diagram that labels the chip loop, facility loop, heat exchangers, chillers, pumps, outdoor equipment, and makeup-water connections. Ask which components operate during ordinary weather, summer peaks, equipment failure, and commissioning. This diagram should be understandable without disclosing security-sensitive control details or proprietary server specifications.
Evaluate the complete operating envelope
The assessment should compare annual electricity, annual water, hottest-day demand, maintenance requirements, and redundancy using the same IT load assumptions.
A design that minimizes one metric may not minimize the others. Promotional comparisons are unhelpful when one case assumes full utilization and another assumes lightly loaded equipment.
Questions that change the conclusion
Is the reported water figure measured, modeled, or a design ceiling? Does “closed loop” refer only to internal circulation? Is occasional evaporative assistance included? Are initial filling, flushing, and treatment water counted? What happens when the primary heat-rejection equipment is unavailable? These questions convert a generic technology label into a usable project assessment. The recommended finding should describe the actual architecture and operating modes, not rank every liquid-cooled or air-cooled facility as environmentally equivalent. Any claimed improvement should retain its baseline, location, load level, and weather assumptions.
Evaporative systems: conserve water without hiding consumption
Makeup and blowdown · Chemistry matters · Meter both sides
Why towers need replacement water
Cooling towers lose water through evaporation and other pathways. Blowdown removes concentrated dissolved solids; replacement makeup water restores the circulating system. EPA guidance recommends controlling chemistry, monitoring conductivity, and measuring makeup and blowdown. Higher cycles of concentration can reduce blowdown, but feasible operation depends on incoming water quality and equipment protection. [54]
Interpret savings correctly
Reducing blowdown is valuable, but it does not mean the same reduction in evaporation or net consumption. A report should identify whether an intervention lowers delivered water, wastewater discharge, evaporative losses, or all three. Avoid presenting a reduction in one stream as proof that the campus has become water independent.
Recommended operating review
Ask for time-aligned meter readings, conductivity records, treatment settings, abnormal-loss investigations, and maintenance logs.
Compare performance with delivered cooling service and weather, not only the previous month. A falling water bill during reduced computing activity is not automatically evidence of better cooling efficiency.
Practical accountability
Set a documented response for unexplained increases in makeup water, persistent overflows, or inconsistent readings. Require contractors to identify whether their performance claim includes all towers and auxiliary equipment. Reconcile utility bills with submeter totals and explain material gaps rather than silently allocating the difference to evaporation.
The preferred public summary is a compact annual balance accompanied by peak-day demand and a plain-language explanation of treatment constraints. Detailed chemical handling belongs with trained operators and the relevant permit authorities. The environmental conclusion should be based on measured flows and demonstrated control, not simply the presence of an automated treatment system or a vendor efficiency guarantee.
Dry and hybrid cooling: avoid a false water-versus-energy choice
Climate-specific tradeoff · Annual plus peak · Test failure modes
Efficiency depends on the site
ENERGY STAR explains that water-side economizers can bypass mechanical chillers when outdoor conditions permit, while DOE recommends climate-specific engineering evaluation of cooling options. “Free cooling” describes reduced mechanical cooling demand; it should not be interpreted as literally zero energy or zero water. [55] [56]
Recommended comparison
Compare candidate systems against one common hourly weather file, one IT load profile, and the same reliability objective. Show electricity and water together for ordinary years and stressed conditions. Include auxiliary pumps and fans, not only the main chiller. Record when a hybrid system switches operating mode and what resource demand that change creates.
Look beyond annual averages
For decision-making, the hottest hours may be more important than the annual mean.
Ask whether the cooling design can maintain required operation during a heat wave without exceeding water commitments or electrical service capacity. Also test the loss of one major cooling component rather than relying solely on the all-equipment-available case.
Recommendation, not a universal ranking
A low-water design may be especially attractive where potable supply is constrained; another site may justify a different combination of water quality, electricity availability, and heat rejection. The report should not assign an automatic environmental winner based on technology name. Ask the engineer to present a transparent tradeoff table and a sensitivity analysis instead.
Before accepting a claimed improvement, verify that it survives plausible changes in climate, server density, and utilization. The evidence should distinguish an operating commitment from a modeled option that has not been purchased, commissioned, or tested. That distinction becomes crucial when phased expansion changes the original cooling assumptions.
Water geography: national totals do not decide local adequacy
Basin before nation · Seasonality matters · Competing commitments
Start with the actual water source
A national consumption estimate describes scale, not whether a local supply can support another campus. Michigan's data-center guidance distinguishes direct water demand and groundwater protections, including review of potential adverse resource impacts under its state framework. That example illustrates why local hydrology and jurisdiction matter; it is not a nationwide permitting rule. [57]
Recommended geographic inventory
Identify the utility service area, intake or aquifer, source watershed, wastewater destination, and any interbasin transfer. Map the project against existing users and committed future withdrawals. A nearby reservoir does not prove that its capacity is available to the project, and a municipal connection does not remove the need to understand the upstream supply.
Stress the supply, not just the campus
Request normal-year, drought-year, and peak-season supply assessments.
Distinguish legal allocation, physical availability, treatment capacity, pipe capacity, and operational reserve. Each can constrain service independently. The recommended review should show who receives priority during shortage and whether campus commitments are enforceable in those conditions.
Fair comparisons
Compare additional demand with the relevant basin and utility system, using the same time interval and flow definition. Avoid comparing annual campus consumption with a reservoir's total stored volume without discussing replenishment and other obligations. Likewise, national agricultural water totals cannot resolve a neighborhood's specific concern about wells or pressure.
The final finding should state the evidence needed to conclude that supply is adequate. Where the source is uncertain or confidentiality prevents verification, classify adequacy as unverified rather than assuming either abundance or depletion. Ask the public utility or responsible resource authority to document the basis for its service commitment.
Scale context: published water estimates and their limits
2023 US estimate · Historical benchmark · Different boundaries
U.S. historical estimate
Berkeley Lab's 2024 report estimates direct U.S. data-center water consumption at approximately 66 billion liters in 2023. It attributes 84 percent to hyperscale and colocation facilities. The same report estimates an indirect electricity-related footprint near 800 billion liters. These are modeled sector estimates, not a census of every site's meters. [58]
A benchmark, not today's meter reading
The report was published in 2024 and describes 2023 historical consumption alongside future scenarios. Keep those categories distinct. A forecast for a later year should never be relabeled as observed consumption just because that year has arrived. For this chapter, the historical estimate is retained as a transparent benchmark rather than presented as a current measured total. [58]
Why the numbers should not be merged
When comparing this benchmark with another study, inspect its year, geography, methodology, and system boundary.
A larger or smaller figure does not by itself establish a growth rate or show that either study is invalid.
This report recommends placing them in separate labeled panels with source edition and observation year visible. A publication date is not the year being measured.
What the figures can establish
They demonstrate that direct facility water is only one part of the issue and that accounting choices materially affect the result. They cannot determine the water demand of an individual proposed campus. Do not divide a national figure by a facility count and present the average as a site's expected consumption.
For project appraisal, use these totals as context and replace generic ratios with engineering forecasts and later meter readings. Preserve uncertainty and model assumptions in the evidence record. Revisit the benchmark when a revised official estimate is available rather than silently mixing editions in a trend line.
Municipal infrastructure: supply agreements are not complete capacity studies
Treatment plus conveyance · Who pays? · Phase-linked commitments
Separate the infrastructure questions
A water-service letter can be important evidence, but the environmental review should ask what it actually commits to deliver. Iowa lists water use, wastewater construction, discharge, stormwater, and floodplain approvals among potentially relevant activities. These are separate resource and infrastructure questions rather than interchangeable assurances. [59]
Recommended utility review
Request the campus's average and maximum daily delivery requirements, fire-flow assumptions, wastewater characteristics, construction demand, and expansion schedule. Match those requirements to treatment, storage, pumping, and distribution capacity. State clearly whether planned upgrades are funded, designed, permitted, or already operating. A future improvement should not appear in the available-capacity column as though it already exists.
Allocate costs and risks explicitly
The recommended financial annex should identify developer contributions, utility investment, connection fees, ongoing charges, and any costs that could fall on other customers.
This is an analytical request, not a finding that a particular tariff subsidizes data centers. Also identify who bears the cost if a project pauses after infrastructure has been enlarged.
Make growth conditional
Tie each expansion phase to verified service availability rather than allowing all phases to inherit an early conceptual letter. Require a clear process for revising forecasts when computing density or cooling architecture changes. Review whether sewer treatment or reuse infrastructure becomes the limiting factor even when raw water remains available.
A strong agreement should explain shortage procedures and monitoring without promising immunity from every emergency. Public reporting can summarize capacity and accountability while protecting sensitive infrastructure details. The central test is whether the proposed operating envelope fits the system that will actually be available when the facility opens.
Wastewater and reclaimed supplies: quality is part of quantity
Source quality · Return destination · No automatic zero impact
Reuse changes the water source
Reclaimed water may reduce reliance on potable supplies, but its suitability depends on treatment, system design, and the receiving environment. Minnesota's data-center oversight explicitly includes cooling-water and wastewater discharges. A supply labeled recycled therefore still needs an account of what enters the site, what leaves, and which authority oversees the discharge. [51]
Recommended evidence
Ask for source-water quality specifications, treatment responsibility, delivery reliability, blowdown destination, and discharge conditions. Determine whether the proposed reuse system is existing infrastructure or a future project requiring new capital and approvals. Separate an executed supply arrangement from a memorandum expressing interest.
Trace the counterfactual
For analysis, ask where reclaimed water would otherwise go. Reusing water that previously supported downstream flows is not equivalent to creating a new resource from nothing.
This does not make reuse undesirable; it means the benefit should be evaluated against the real alternative rather than a slogan.
Operating assurance
Request a contingency for periods when reclaimed supplies fail to meet specification. Does the campus switch to potable water, reduce activity, store enough water, or use a different cooling mode? Quantify that contingency in the supply assessment. A low annual potable share can conceal a large peak call on drinking-water infrastructure.
The recommended public metric set includes potable deliveries, reclaimed deliveries, consumption estimate, discharge volume, and significant quality exceptions. Avoid disclosing unnecessary treatment-security detail. A defensible conclusion explains both the avoided potable demand and the remaining water-management obligations. It does not equate reuse, recycling, replenishment, and consumption reduction as though those terms describe the same physical result.
Electricity-related emissions: average footprint versus added load
Location matters · Average ≠ marginal · Preserve assumptions
Two legitimate but different calculations
EPA's eGRID provides generation, emissions, and emission rates across several geographic aggregations. Its guidance distinguishes annual output rates from nonbaseload rates used in certain avoided-emissions analyses. An inventory describing electricity consumption and a study asking what additional load causes are not necessarily using the same emissions factor. [60]
Recommended reporting practice
Label whether a calculation is historical attribution, prospective system modeling, or a scenario sensitivity. Record electricity quantity, year, grid region, emissions factor, gases included, and treatment of losses. Do not call a national average a measurement of the power plants physically responding to one campus.
An illustrative sensitivity test
At 100,000 megawatt-hours, a factor of 0.2 metric tons per megawatt-hour produces 20,000 metric tons; a factor of 0.6 produces 60,000.
These are illustrative assumptions, not observed site emissions. The arithmetic shows why regional and methodological choices can dominate a result.
Interpret procurement separately
The recommended assessment should show physical electricity consumption alongside any contractual claims, with their accounting basis clearly identified. Purchasing clean-energy instruments does not by itself document the generating mix in every hour of operation. Conversely, a location-based estimate does not capture every investment effect of a procurement contract.
For a proposed facility, ask a qualified system analyst to test alternative generation and transmission scenarios. Report the range and its drivers rather than manufacturing a single precise forecast. The most useful conclusion explains which decisions could lower the footprint and what evidence would demonstrate that improvement after operation begins.
On-site power: classify the operating role before estimating pollution
Backup versus primary · Fuel and runtime · Permit-specific limits
The same equipment can serve different roles
EPA's data-center Clean Air Act resource identifies stationary engines and turbines as potential primary and backup sources subject to applicable emissions standards. The environmental consequences depend on fuel, controls, load, operating time, and location, not simply the number of generators on a plan. [61]
Recommended equipment register
For each source group, record purpose, rated output, fuel, emission-control design, anticipated operating hours, and the governing permit. Distinguish emergency operation, readiness testing, temporary construction supply, and regular generation. Do not assume a device called backup equipment operates under an emergency-only authorization.
Evaluate realistic operating cases
The recommended emissions assessment should include ordinary testing, outage operation, and any planned nonemergency use. Consider simultaneous operation across the campus where relevant.
Annual totals and short-duration concentrations answer different questions, so one should not substitute for the other.
Separate proposed and demonstrated control
Ask whether control equipment has been specified, installed, commissioned, and verified. Require operating records and explain how maintenance or failure affects compliance. A vendor emissions claim at one test condition is not a complete site assessment.
Public discussion should avoid two opposite errors: treating permitted maximum hours as actual historic operation, or treating low average runtime as proof that peak events are immaterial. The defensible approach presents actual records where available and explicitly labeled bounding scenarios where they are not. Changes from emergency to routine generation should trigger a fresh review of the project's environmental narrative, infrastructure needs, and applicable approvals rather than being treated as a minor descriptive update.
Local air quality: measure the pathway to receptors
Emissions ≠ exposure · Cumulative context · Independent review
Source totals are only the beginning
The Bay Area Air District explains that its data-center role focuses on stationary pollution sources, predominantly backup generators, rather than every aspect of the facility. Its published process includes source permitting, control evaluation, and operating requirements. This is a jurisdiction-specific example of why an air permit should be read for its actual scope. [62]
Recommended technical questions
Ask which pollutants are modeled, what background concentrations are used, which receptors are evaluated, and how meteorology and nearby sources are represented. Confirm that the modeled equipment layout matches the current design. A changed stack location or larger generator fleet can make an old illustration a poor description of the proposal.
Keep health claims proportional to evidence
An emissions inventory does not by itself establish personal exposure or diagnose a health effect.
The recommended report should distinguish release estimates, modeled concentrations, monitoring observations, and health interpretation. Qualified specialists should address health-risk methods; public summaries should explain the limitations without dismissing concerns.
A practical verification package
Request the permit application, modeling report, regulator response, final conditions, and later operating records. Summarize the applicable comparison standard and whether the result is a prediction or observation. Include a plan for investigating complaints and unusual events, with responsibilities assigned before commissioning.
Cumulative assessment should identify the relevant airshed and nearby sources instead of drawing an arbitrary circle and assuming every emission has equal effect. Where data are absent, identify the specific missing input. The objective is not to imply that every campus creates the same exposure, but to establish an auditable path from equipment and operation to the conditions experienced by nearby people.
Noise: continuous operation needs more than a single reading
Baseline conditions · Tonal character · Post-opening verification
Why the operating pattern matters
Virginia's JLARC review identifies the constant nature of data-center noise as a problem in some locations near residences. Its findings also highlight land-use compatibility. This is evidence that nuisance can be site-specific, not a basis for claiming that every data center exceeds a universal noise limit. [63]
Recommended baseline
Before construction, document representative daytime and nighttime conditions at relevant receptors. Record weather, measurement duration, equipment, and unusual background events. Preserve enough detail for a later comparison. A reading beside a busy road should not stand in for a quieter residential receptor without explanation.
Model the complete campus
The recommended acoustical study should include cooling equipment, electrical equipment, generators where relevant, and future phases.
Ask how tonal or low-frequency characteristics are addressed, how nighttime conditions are evaluated, and whether barriers and building orientation are included in the model. Local standards and measurement methods must be verified rather than borrowed from another jurisdiction.
Verify after opening
Require testing under a defined operating condition, not simply whenever a convenient measurement can be scheduled. Record the equipment actually running during the test. Establish complaint intake, investigation timelines, and corrective-action responsibility. Visual landscaping may improve appearance but should not be accepted as demonstrated sound control without engineering support.
The public report should present compliance findings separately from unresolved nuisance concerns. A short daytime survey cannot establish every nighttime condition, and a phone recording is not equivalent to a calibrated compliance study. Both resident observations and technical measurements can inform investigation when their roles and limits are clearly stated.
Land, habitat, and off-site corridors: assess the whole development
Footprint versus parcel · Avoidance first · Include utility corridors
The parcel is not the impact boundary
The assessment should distinguish owned acreage, disturbed acreage, impervious area, protected buffers, and supporting corridors. A campus may require water mains, substations, transmission connections, roads, and drainage works beyond its fence. EPA's Section 404 overview describes review of dredged or fill discharges into covered waters, including wetlands, with avoidance, minimization, and compensation as key considerations. Applicability requires project-specific review. [64]
Recommended mapping
Overlay the current site plan with field-verified streams and wetlands, habitat surveys where warranted, slopes, existing vegetation, cultural resources, and neighboring uses. Desktop screening is a starting point, not a substitute for the required field or agency determinations.
Test alternatives before mitigation
For analysis, compare different building placements, road alignments, stormwater layouts, and infrastructure routes before assuming every disturbance is unavoidable.
Record why alternatives were rejected. A compact building footprint can still create substantial off-site effects if its utility corridors cross sensitive areas.
Keep accounting transparent
Report temporary disturbance separately from permanent conversion. Identify who maintains buffers and replacement planting, how survival is checked, and what happens if promised mitigation fails. Do not treat planted acreage as automatically equivalent to mature habitat lost elsewhere.
The recommended decision package includes a map of all phases and a disturbance ledger that changes with design revisions. Photographs should carry dates and locations so they can document conditions rather than merely illustrate a preferred narrative. The final assessment should acknowledge tradeoffs between existing land uses, development benefits, and ecological constraints without inventing a universal acreage-per-megawatt benchmark. Site design and surrounding geography determine much of the actual impact.
Construction: temporary impacts can last through multiple phases
Dust and traffic · Stormwater controls · Cumulative schedule
Construction is a distinct operating period
Clearing, grading, excavation, stockpiles, equipment movement, and deliveries create a different impact profile from an occupied data hall. EPA's construction stormwater guidance covers qualifying disturbances of one acre or more, including smaller sites within a larger common development plan. The applicable permitting authority and current permit conditions still need to be confirmed for each project. [65]
Recommended construction ledger
Identify the disturbed area, work sequence, haul routes, stockpile locations, drainage controls, working hours, and responsible contractors. Include utility and road construction supporting the campus, not only the building contractor's fenced area. Separate a single building's duration from the full campus schedule.
Plan for overlapping phases
A facility may begin operations while additional buildings remain under construction. For assessment, combine operational traffic and equipment with continuing construction activity.
Require a process for updating the public schedule when work slips or overlaps. A claim that impacts are temporary should state an expected period and acknowledge uncertainty.
Verification in the field
Recommended evidence includes inspection logs, corrective-action closure, dated photographs, approved traffic plans, and complaint records. After major weather events, review whether controls performed as intended. Track repeated failures rather than counting every inspection as a successful outcome.
Dust, vibration, access, and noise concerns should be assigned to the appropriate authority or contract condition. An environmental discharge permit does not necessarily regulate every neighborhood concern. The strongest construction plan combines enforceable requirements with a reachable site contact and clear escalation path. This is a recommended management framework, not a claim that every listed measure is universally mandated by law.
Flooding and drainage: maps are a baseline, not a guarantee
Pluvial and river risk · Downstream effects · Critical equipment elevations
Screen broadly, then investigate
Massachusetts' official guidance explains limitations in using Flood Insurance Rate Maps and Flood Insurance Studies as the sole estimate of flood risk. That distinction is relevant to any infrastructure screening: a regulatory map is useful evidence, but it does not answer every question about rainfall, drainage, future conditions, or site changes. [66]
Recommended site review
Request current mapping, topographic survey, drainage calculations, historical flood information, and proposed grading. Identify overland flow paths and downstream receptors. Examine access roads and utility connections as well as buildings. A protected data hall can still lose support services if surrounding infrastructure floods.
Avoid transferring risk
For analysis, compare predevelopment and proposed runoff characteristics using an accepted local method. Ask how detention, infiltration, conveyance, and emergency overflow are designed and maintained.
Do not infer that a pond visible on a rendering proves adequate stormwater management.
Verify the as-built condition
Recommended documentation includes surveyed elevations, constructed drainage features, maintenance responsibility, inspection access, and contingency procedures. Check whether later phases alter the assumptions of an earlier drainage study. Fuel systems and other environmentally sensitive equipment deserve explicit attention in the flood assessment.
The final report should state the evaluated event, model assumptions, and residual uncertainty. “Outside the mapped floodplain” is not equivalent to “cannot flood,” and a past absence of flooding is not a complete forecast. Conversely, a broad hazard screen does not establish that a particular engineered design is unsafe. The appropriate conclusion follows from site-specific evidence and an explicit review of both campus resilience and effects on neighboring properties.
Fuel, refrigerants, and equipment: inventory the less visible impacts
Storage pathways · Leakage records · End-of-life evidence
Resources beyond electricity and water
Backup systems and cooling equipment introduce materials that need their own controls. EPA's oil-storage compliance guidance describes SPCC applicability based on capacity and potential discharge pathways, with conditions and exclusions that require site review. EPA also publishes changing HFC sector restrictions relevant to cooling equipment. Do not substitute an old summary for the current rule and equipment classification. [67] [68]
Recommended inventory
List fuel storage, major cooling-fluid and refrigerant systems, batteries, and other significant managed materials. Record responsible operators, inspection requirements, incident procedures, and disposal routes. Protect detailed security-sensitive layouts while making the accountability framework available.
Track actual losses and destinations
The recommended environmental ledger includes purchases, major top-ups, recoveries, transfers, and documented disposal. Repeated replenishment without explanation deserves investigation.
A maintenance contract alone does not establish that leakage is negligible or that waste reaches an appropriate destination.
Equipment retirement
EPA recommends certified electronics recyclers for unwanted electronics. The useful evidence is the actual downstream management arrangement, not a generic claim that equipment is recyclable. [69] Ask for chain-of-custody records and distinguish reuse, refurbishment, material recovery, and disposal. Include secure data handling in the operational plan without confusing it with environmental certification.
This page identifies review categories rather than prescribing chemical handling or emergency response. Qualified personnel and relevant authorities should determine technical controls. The final assessment should explain which materials are present, how releases are prevented, and what evidence supports responsible retirement. It should not assume that a facility with low operating emissions necessarily has an immaterial materials footprint.
Embodied impacts: construction and replacement belong in the ledger
Materials boundary · Comparable declarations · Avoid double counting
The footprint begins before operation
Concrete, steel, electrical equipment, servers, transport, and construction activities belong in a broader lifecycle assessment. EPA's sustainable materials guidance emphasizes treating construction and demolition materials as recoverable resources where appropriate, reducing demand for virgin materials. That principle supports lifecycle evaluation, but it does not provide a universal embodied-carbon factor for a data center. [70]
Recommended procurement evidence
Request material quantities, product-specific environmental declarations where available, transport assumptions, and the lifecycle stages included. Compare alternatives using consistent functional performance and accounting boundaries. A lower figure is not meaningful if one product excludes stages included by another.
Handle replacement honestly
For analysis, identify assumed service lives for the building shell, mechanical plant, electrical systems, and computing equipment. Test how replacement cycles affect the result rather than assigning the entire campus one lifetime.
Do not count a purchase in multiple categories simply because both the owner and a tenant report it.
Separate certainty levels
Classify quantities as tender estimates, installed quantities, or documented retirements. Record generic factors separately from supplier-specific data. If an environmental declaration covers a different plant, product, or production period, note the mismatch rather than treating it as a direct measurement.
Recommended public reporting should show embodied and operating impacts separately before combining them. This avoids hiding a construction footprint behind an electricity-procurement claim or implying that one efficient year erases all prior impacts. The useful decision is comparative: which feasible design or procurement change reduces lifecycle burden while meeting structural, safety, reliability, and service requirements? Any savings estimate should retain the alternative baseline and the uncertainty that comes with it.
Heat reuse: value only the heat that has a real user
Temperature and distance · Seasonal demand · Delivered heat, not potential
A physical opportunity, not automatic credit
DOE's design guidance explains that useful heat reuse depends on a nearby heat user and a suitable temperature level; backup cooling generally remains necessary when that user is unavailable. Its guidance favors evaluating actual integration conditions rather than assuming all rejected heat becomes a useful product. [56]
Recommended feasibility test
Identify the heat customer, demand profile, delivery temperature, distance, pipe route, and any heat-pump requirement. Match hourly or seasonal demand with campus output. A cold-season heating market may not accept the same quantity in summer. Ask who funds and operates the connection and who takes risk if either party changes operations.
Count delivery, not a rendering
For reporting, distinguish recoverable heat potential, installed recovery capacity, heat exported, and heat actually used to replace another energy input.
The EU data-center reporting framework defines reused energy in terms of heat used outside the boundary that substitutes for outside energy needs. That is a useful accounting reference, although its legal applicability is jurisdiction-specific. [71]
Evaluate the net result
The recommended calculation includes additional pumping, heat-pump electricity, distribution losses, and the displaced heating baseline. Do not claim the same recovered heat twice through different credit categories. Nor should a heat-reuse plan justify inefficient computing simply because more wasted energy could be exported.
A credible project has a signed operational arrangement, a commissioned connection, measured delivery, and a clear method for estimating displacement. Before those exist, describe the proposal as potential rather than achieved benefit. The environmental conclusion should explain both the useful energy recovered and the remaining need for heat rejection when the external customer cannot receive it.
Verification: turn environmental promises into an auditable record
Evidence register · Phase gates · Publish exceptions
Establish who regulates what
Mississippi's data-center guidance explicitly separates environmental permitting from issues such as zoning, traffic, noise, and utility rates handled through other processes. Tennessee likewise describes equipment-focused air regulation rather than a single comprehensive approval of the data center. The practical lesson is to build a responsibility matrix, not assume one permit resolves every concern. [72] [73]
Recommended evidence register
For each material claim, record its author, date, boundary, unit, status, source document, and verification owner. Use distinct labels for proposed, permitted, constructed, commissioned, measured, and independently reviewed. Preserve superseded versions so a changed design cannot quietly inherit evidence from a different project.
Set review gates
Before site approval, seek a credible impact baseline and alternatives assessment. Before construction, verify the required permissions and management plans.
Before operation, reconcile the built equipment with the approved design. During operation, compare actual water, energy, generation, complaints, and incidents with commitments. These are recommended governance gates, not a statement that every jurisdiction requires an identical process.
Report what went wrong as well as what improved
An annual dashboard should disclose material exceptions, corrective action, unresolved uncertainty, and expansion changes alongside performance metrics. Explain missing data rather than filling gaps with unmarked estimates. Where independent review is promised, specify its scope and whether findings are public.
The final environmental judgment should be conditional on evidence: what is known, what remains uncertain, which mitigations are demonstrated, and what would trigger reconsideration. This supports accountable development without treating every campus as either inherently harmless or inherently unacceptable. Project-specific measurements, transparent boundaries, and enforceable responsibilities are more informative than a sustainability label alone.
Exhibit 11 · Water use extends beyond the site
IEA estimated 885 billion L of total global data-center-associated water consumption in 2025, of which 240 billion L was direct. Other associated use (645 billion L) is a residual, including electricity generation, manufacturing, and primary energy. Do not compare these totals with a local withdrawal permit.
IEA, World Energy Outlook 2026, Figure 1.39: primary source
Exhibit 12 · Hyperscale estimates have a range
[*REDACTED*] and colleagues modeled 403 U.S. hyperscale data centers operating May 2024 to April 2025. The intervals shown are facility-load scenarios, not statistical confidence limits. Their central location-based emissions estimate was 25.7 Mt CO₂; scenario range about 21–31 Mt.
[*REDACTED*] et al., Nature Sustainability, November 16, 2026: primary source
Exhibit 13 · Cooling and water accounting
Conceptual site and upstream boundaries. Withdrawal records, return flows, and consumption records answer different licensing and community questions.
48PEGASUS Research original conceptual exhibit. Not empirical data.
Browse: White Papers · Reports · Downloads
