How to Use This Dashboard

Click pathway cards throughout this dashboard to build a custom summary of exposure pathways relevant to your context. Use the sidebar to navigate lifecycle phases; selected pathways can be exported or printed for further review. Cards reflect pathways described in the literature; this tool is a reference resource, not a regulatory instrument or establishing a causal relationship.

What Are Data Centers?

Data centers are large-scale facilities that house computing infrastructure — including servers, networking equipment, and storage systems — supporting cloud services, artificial intelligence operations, and digital communications.2 The United States hosts approximately half of the world's data center facilities.1 Facilities vary substantially in scale, from small localized systems to hyperscale campuses, and require substantial electricity for computing and cooling, significant water use depending on cooling system design, and supporting infrastructure including backup power systems and transmission capacity.2,7
Box A: Data Center Types
Micro Data CenterFewer than 10 server racks. Continuous power draw: under 150 kW.
Small Data Center5,000–20,000 sq ft; 500–2,000 servers. Continuous power draw: 1–5 MW.
Average Data Center20,000–100,000 sq ft; 2,000–5,000 servers. Continuous power draw: ~100 MW.
Hyperscale Data Center100,000+ sq ft; 5,000+ servers. Continuous power draw: over 100 MW.
Power draw is measured in kilowatts (kW) or megawatts (MW) — units of power, meaning the rate of electricity consumed at any given moment, continuously, around the clock. 1 MW = 1,000 kW. To put it in context: an average U.S. home uses roughly 1.2 kW continuously. These figures represent the facility's constant baseline electrical demand, not daily or annual totals.
Source: IBM (2026). ibm.com

About This Dashboard

This dashboard synthesizes findings from a scoping review — Armstrong, NS, Cullen, T, Neatherlin, J & Exum, NG (2026). "Direct and Indirect Public Health Impacts of Data Center Development: A Scoping Review." Johns Hopkins Bloomberg School of Public Health, Dept. of Environmental Health and Engineering. Prepared for Pima County Health Department — to support health departments and other agencies in identifying and communicating exposure pathways and potential health effects across the complete lifecycle of data center development. The review identified 499 sources; following two-stage PRISMA-aligned screening, 80 met inclusion criteria. The dashboard was co-designed with public health practitioners and draws on the EPIS (Exploration, Preparation, Implementation, Sustainment) implementation science framework,9 the ASTHO/NACCHO Health in All Policies evaluation tool,10 and the CDC 10 Essential Public Health Services.11

No empirical epidemiologic studies directly measuring health outcomes in populations residing adjacent to data centers were identified in the scoping review. Available evidence consists predominantly of exposure modeling studies (30%), theoretical reviews (45%), policy analyses (15%), and environmental monitoring studies (10%). All quantitative projections referenced in this dashboard derive from modeling and inferential analysis rather than measured population health outcomes. This dashboard communicates potential pathways consistent with available literature; it does not establish causal relationships or provide regulatory thresholds.

Evidence Landscape

  • Theoretical or review papers (45%, ~36 sources): Synthesizing existing evidence without primary data collection.
  • Exposure modeling studies (30%, ~24 sources): Projecting community-level exposures using mathematical models.
  • Policy analyses (15%, ~12 sources): Characterizing regulatory approaches and recommendations.
  • Environmental monitoring studies (10%, ~8 sources): Measuring environmental conditions near facilities.
  • Empirical epidemiologic studies: None identified. No studies directly measuring health outcomes in data center-adjacent populations were found in the scoping review.
0
Initial
Classification & data
1
Pre-Siting
Equity, air, water, land
2
Construction
Diesel, dust, noise
3
Operations
Air, water, heat, noise
4
End-of-Life
E-waste, PFAS, batteries

This map displays 1,895 U.S. data center clusters alongside the cropland, grid, population, and water context of each site. Use the search function to look up an existing, under construction, proposed, or canceled facility, or use the Check a New Location tool to enter an address or select an area to explore how a potential site interacts with food, energy, and water systems.

Open full map ↗ Data and methods ↗
0
Phase 0
Initial Considerations: Facility Characteristics

Pathway Inquiry Legend

Pathway cards throughout this dashboard are interactive — click any card to select it for your summary. Selection indicates pathways described in the literature that may be considered for further examination within the local context. No card is pre-selected; this is a structured reference tool, not a regulatory instrument. Evidence type (measured, modeled, or inferred) is noted on each card.

Facility Size Classification

Data centers are classified by power demand and physical scale, ranging from edge facilities serving localized needs to hyperscale campuses supporting global cloud and AI infrastructure.7 Facility scale influences the magnitude of environmental demands across all lifecycle phases.

Size ClassPower DrawApproximate FootprintTypical Scale
Micro / EdgeUnder 150 kilowatts (continuous)<5,000 sq ft<10 server racks
Small1–5 megawatts (continuous)5,000–20,000 sq ft500–2,000 servers
Mid-Tier10–100 megawatts (continuous)20,000–100,000 sq ft2,000–5,000 servers
HyperscaleOver 100 megawatts (continuous)≥100,000 sq ft≥5,000 servers

Cooling System Classification

Cooling system design can shape the distribution of environmental burdens across scopes.12,13 Understanding the tradeoffs between cooling types is important because water-cooled systems consume more direct water on-site, whereas air-cooled systems require more electricity, shifting burden to upstream power generation facilities.12

Scope 1 — Direct On-Site Resource UseResource consumption and emissions occurring directly at the facility, including cooling water withdrawals, diesel generator exhaust, and refrigerant handling. Scope 1 impacts are most proximate to fenceline communities.
Scope 2 — Indirect Upstream (Electricity & Power Plants)Impacts mediated through electricity consumption — including air emissions and water withdrawals at power generation facilities — which may affect communities geographically distant from the data center.8,14 supply-chain manufacturing impacts (Scope 3) extend further upstream.15
Scope 3 — Supply-Chain ManufacturingImpacts embedded in the manufacture of servers, chips, and hardware before the facility is built or operated. Scope 3 is the hardest to quantify but can represent substantial embedded water and carbon. It is not typically reported by operators.
Cooling TypeDirect On-Site Water Use (Scope 1)Upstream Electricity Demand (Scope 2)Key Consideration
Air-Cooled Very Low High Shifts water burden upstream; increased electricity demand may intensify air quality and water use at power generation facilities.12
Evaporative Cooling High Moderate Direct local water withdrawal is a primary community-level consideration; lower electricity demand relative to air cooling may reduce upstream electricity burden (Scope 2).5,13
Liquid / Immersion Cooling Moderate Low–Moderate Generally more efficient across scopes; newer technology with a more limited long-term evidence base.
Hybrid Systems Variable Variable Hybrid cooling systems represent an emerging technology currently being implemented in data centers; long-term operational data are limited.

Suggested Information to Request

When available, requests for information may include: electricity demand and source attribution; cooling system type and direct on-site water use (Scope 1); estimated Scope 2 impacts where available; backup generator specifications (capacity, test frequency, estimated NOx). Transparency in environmental performance data is noted in the literature as a persistent barrier to health assessment, as non-disclosure agreements and absence of mandatory reporting requirements have constrained assessment capacity.16

1
Phase 1
Pre-Siting Considerations

Six parallel pathways establish the baseline environmental and community health context for a proposed site.

Click any pathway card to add it to your summary for further review. Cards can be deselected by clicking again.
✓
Pathway 1A
Environmental Equity Screening

Data centers in the U.S. can be sited in areas with elevated pre-existing environmental burden; incremental health impacts attributable to data center siting have not been directly measured at the time of this review.

Existing indices provide a starting point for characterizing whether proposed sites are located in areas with pre-existing environmental burden or social vulnerability. Validated indices including EPA EJScreen, CDC Social Vulnerability Index (SVI), and the U.S. Climate Vulnerability Index characterize pre-existing environmental burden and social vulnerability in proposed site areas.23 Modeled analyses suggest that per-household health burdens in disadvantaged communities near data center-associated infrastructure may be substantially greater than in less-impacted areas.17,18
Example tools: EPA EJScreen · CDC SVI · U.S. Climate Vulnerability Index · State EE Mapping Tools
Evidence basis: Modeled equity analyses; no direct epidemiologic measurement of data center-specific equity burden in exposed communities.17,18
✓
Pathway 1B
Regional Baseline Air Quality

Approximately 95% of U.S. data centers are located in grid regions where electricity carbon intensity exceeds the national average; health outcomes attributable to data center electricity demand have not been directly measured at the time of this review.

Regional ambient air quality conditions establish the baseline against which facility-related changes can be contextualized. Ambient PM2.5 and ozone concentrations, assessed against National Ambient Air Quality Standards (NAAQS), characterize baseline respiratory burden in the region prior to facility development.14 Approximately 95% of U.S. data centers are located in areas with electricity carbon intensity exceeding the national average, contributing to upstream power plant air (Scope 2) quality implications from grid-level electricity demand.14 When appropriate, grid carbon intensity can be estimated using tools such as EPA eGRID to contextualize the air quality considerations associated with a facility's projected electricity demand.14
Example tools: EPA AQS · EPA eGRID · AirNow · CDC EPHTN
Evidence basis: Exposure modeling and environmental monitoring; ~95% carbon intensity finding from Guidi et al. (2024).14
✓
Pathway 1C
Water Stress

Since 2022, approximately 40% of U.S. data centers have been located in water-stressed regions; associations between data center water consumption and community health outcomes have not been directly measured at the time of this review.

Depending on facility specifications and cooling system design, data centers use direct and indirect water across multiple scopes for cooling and electricity generation. Regional freshwater availability relevant to proposed facility siting can be characterized using water stress indices such as the WRI Aqueduct Water Risk Atlas.13 Approximately 40% of existing U.S. data centers are located in water-stressed areas, including regions of the Southwest, Midwest, Southeast, and Southern Plains.13 Comprehensive water assessment requires accounting across direct on-site cooling (Scope 1), Scope 2 (power plant cooling), and supply-chain manufacturing (Scope 3) to capture the full demand profile.15 Downstream impacts may include competition for potable freshwater supplies and agricultural impacts in water-stressed regions.13,15
Example tools: WRI Aqueduct · EPA eGRID · State utility data
Evidence basis: Lifecycle assessments and resource modeling; Chen & Wemhoff (2022), Western Resource Advocates (2025), Li et al. (2025).13,15
✓
Pathway 1D
Land Use & Green Space

Data center development can involve the conversion of green space, cropland, or arid land to impervious surface; health effects attributable to this land conversion in adjacent populations have not been directly measured at the time of this review.

Data center development may involve large-scale land conversion with potential implications for vegetation cover, stormwater management, and community access to green space. Data center development typically involves conversion of agricultural land, forested areas, or undeveloped parcels to impervious surface.19 Satellite monitoring near data center clusters has documented vegetation decline (NDVI reduction) and nighttime light intensity increases associated with infrastructure expansion.20 Green space loss may be associated with adverse health outcomes in other infrastructure contexts — including reduced physical activity opportunity, elevated heat exposure, degraded air quality, and mental health impacts — though these relationships have not been characterized in data center-adjacent populations specifically.21,22
Example tools: NLCD · USGS Landsat · FEMA FIRM
Evidence basis: Satellite environmental monitoring (Chakraborty, 2024; Virmani, 2024); analogous green space health literature.20,21
✓
Pathway 1F
Mental Health

The data center siting process and associated land use changes may be associated with psychosocial stress in nearby communities; this has not been directly studied in data center contexts at the time of this review.

In analogous industrial facility siting contexts, limited opportunities for public participation, perceived lack of transparency in decision-making, and a sense of reduced community control over local land use have been described in relation to psychosocial stress in nearby populations.45 The conversion of familiar landscapes — including agricultural land, wooded areas, or community-identified open space — to large-scale industrial infrastructure has been described in relation to place-based psychosocial stress in communities experiencing involuntary environmental transformation, a phenomenon characterized in the environmental health literature as solastalgia.46 Perceived uncertainty about potential health and quality-of-life impacts from a proposed facility may additionally contribute to anticipatory psychosocial stress; these dimensions of the siting process have not been characterized in data center-adjacent communities specifically.45,46
Evidence basis: Analogous industrial facility siting and environmental change literature; Downey & Van Willigen (2005); Albrecht et al. (2007).45,46
✓
Pathway 1E
Cumulative Impact & Community Vulnerability

Pre-existing environmental burden and health vulnerability may modify community response to additional stressors; this has not been directly measured in data center contexts at the time of this review.

Integration of pathways 1A–1D with existing community health status — including asthma prevalence, cardiovascular disease burden, and socioeconomic vulnerability — provides cumulative context for a proposed facility.23 Communities with existing health burden and environmental stress may experience compounding effects from additional environmental pressures, consistent with cumulative impact frameworks.23
Reference frameworks: NASEM Cumulative Impact Assessment Report (2025)23 · EPA Cumulative Impacts Research: epa.gov/healthresearch/cumulative-impacts-research
Evidence basis: NASEM Cumulative Impact Assessment Report (2025)23; EPA Cumulative Impacts Research: epa.gov/cumulative-impacts.

Pre-Siting: Considerations

Pre-siting is an appropriate stage to gather baseline data and characterize the environmental and health context of a proposed site. Activities at this stage may include: baseline environmental characterization; input into environmental review processes; identification of populations warranting particular consideration; and documentation of existing surveillance indicators.10,11 Findings from this phase can inform which operational pathways (Phase 3) are most relevant to explore for a given facility and regional context.

2
Phase 2
Construction Phase Considerations

Construction of data center facilities typically spans 18–24 months and, in analogous literature, may involve modeled or observed acute environmental exposures — including diesel emissions, fugitive dust, and noise — for communities near the facility boundary.24 Construction is among the less-studied lifecycle phases in the available literature; findings are largely inferred from analogous infrastructure contexts.

Click any pathway card to add it to your summary for further review.
✓
Pathway 2A
Construction Air Quality

Construction of large facilities generates diesel PM2.5, NOx, and fugitive particulate matter; health outcomes in adjacent populations have not been directly measured at the time of this review.

Construction activity may generate diesel exhaust and fugitive dust emissions that could affect air quality near the facility boundary, based on analogous infrastructure literature. Heavy construction equipment — including cranes, excavators, concrete mixers, and loaders — generates diesel PM2.5 and NOx emissions, which have been described in relation to respiratory and cardiovascular health outcomes in other infrastructure contexts.27,24 Excavation and grading activities can produce fugitive particulate matter.24 Communities with pre-existing respiratory burden or proximity to facility boundaries may be of particular relevance.27
Evidence basis: Air quality and GHG emissions assessment of data centers in Texas (Eslami, 2025) and life cycle assessment of a UK data centre (Whitehead et al., 2015).27,24
✓
Pathway 2B
Construction Noise

Data center construction can generate prolonged noise exposure over approximately 18–24 months; noise-attributable health outcomes in adjacent populations have not been directly measured at the time of this review.

In analogous infrastructure literature, prolonged construction noise exposure (18–24 months) may be associated with sleep disturbance, annoyance, and cardiovascular effects in other exposure contexts; chronic environmental noise has been described in relation to hypertension and ischemic heart disease in the epidemiologic literature.25,26 Nighttime construction warrants particular attention given described relationships between nocturnal noise and sleep architecture disruption in analogous contexts.25
Evidence basis: Analogous noise-health literature (Hahad et al., 2023; Zaman et al., 2022); no data center-specific construction studies identified.25,26
✓
Pathway 2C
Traffic & Transport Corridor Exposures

Construction-phase heavy vehicle traffic may alter PM2.5 and NOx exposures along designated haul routes; health outcomes in corridor communities have not been directly measured at the time of this review.

Heavy vehicle traffic during construction may alter regular traffic patterns and generate PM2.5 and NOx exposures along transport corridors in communities beyond the facility boundary. Construction of large facilities requires substantial movement of materials and equipment, and designated haul routes may redirect heavy vehicle traffic through areas that do not typically carry this load.27,24 These altered corridor patterns may affect community exposures along routes that differ from those associated with the facility's permanent operational footprint.27,24
Evidence basis: Air quality and GHG emissions assessment of data centers in Texas (Eslami, 2025) and life cycle assessment of a UK data centre (Whitehead et al., 2015).27,24

Construction Phase Considerations

Health departments may consider requesting a Construction Environmental Management Plan (or similar document) that addresses emission controls, dust suppression, noise monitoring, and community notification procedures.

3
Phase 3
Operational Phase Considerations
Important framing: Operational impacts have been the most extensively studied lifecycle phase, primarily through exposure modeling and environmental monitoring rather than direct measurement of health outcomes.2 Multiple affected populations may be relevant — fenceline communities (Scope 1 exposures), communities near power plants (Scope 2), and transport corridor communities — who may be geographically distant from each other and from the data center itself.14 Understanding which community is potentially affected by which pathway is an important consideration when assessing data center development or operations.
Click any pathway card to add it to your summary for further review.
✓
Pathway 3A — Scope 1
On-Site Air: Diesel Backup Generator Emissions

On-site backup generators are a source of localized NOx and PM2.5 during testing and emergency operation; community-level health outcomes attributable to generator emissions have not been directly measured at the time of this review.

Backup diesel generators are a potential localized source of NOx and PM2.5 during routine testing and emergency operation, with modeled analyses suggesting contributions to regional air quality burden. Generators are typically tested monthly or quarterly and activated during grid outages.27 In Texas, generator testing has been estimated to emit approximately 12 metric tons of NOx annually per large facility, with potential contributions to metropolitan ozone levels.27
Example tools: EPA AQS · AirNow · Fenceline PM2.5 sensors for large facilities
Evidence basis: Exposure modeling (Han et al., 2025; Eslami, 2025); Texas NOx data from Eslami (2025). No measured community health outcomes identified.18,27
✓
Pathway 3B — Scope 2
Upstream Air: Power Plant Emissions

Data center electricity consumption may contribute to upstream air emissions at power generation facilities; projected health impacts are model-derived and have not been directly measured at the time of this review.

Electricity demand from data centers may contribute to air emissions at upstream power generation facilities, affecting communities near those plants rather than communities adjacent to the data center itself. Flow-tracing analysis suggests electricity imports can produce modeled health impacts in regions distant from the consuming facility, with communities near coal-fired and natural gas-fired power plants experiencing the highest modeled pollutant concentrations.14,28
Example tools: EPA eGRID · EPA AQS · Electricity source attribution data
Evidence basis: Exposure modeling (Guidi et al., 2024; Han et al., 2025; Hennessy et al., 2022). Projections not validated against measured health outcomes.14,18,28
✓
Pathway 3C
Water Use (Scope 1 / 2 / 3)

Data centers consume direct and indirect water across multiple operational scopes; associations with downstream community health outcomes have not been directly measured at the time of this review.

Depending on cooling system design, data centers use direct and indirect water across multiple scopes, and full accounting requires looking beyond on-site consumption alone. Water use documentation in the literature typically captures only direct on-site water use (Scope 1) while excluding upstream power plant water use (Scope 2) and supply-chain manufacturing water (Scope 3).15 U.S. data centers are estimated to consume approximately 17 billion gallons annually for direct on-site cooling (Scope 1), with an additional 211 billion gallons consumed through electricity production (Scope 2).29 Approximately 40% of facilities are located in water-stressed regions.13 Downstream impacts may include competition for potable freshwater supplies and agricultural impacts in water-stressed regions.5,15
Example tools: WRI Aqueduct · EPA eGRID · State SDWIS · Utility data
Evidence basis: Lifecycle and resource modeling (Li et al., 2025; Western Resource Advocates, 2025; Kollar & Grady, 2025).13,15,29
✓
Pathway 3D
Operational Noise

Data centers generate continuous operational noise from HVAC and power distribution infrastructure; noise-attributable health outcomes in adjacent populations have not been directly measured at the time of this review.

Continuous operational noise from data center infrastructure has been documented at elevated levels in nearby residential areas, and chronic noise exposure may be associated with cardiovascular and sleep outcomes in analogous contexts. Data centers generate continuous operational noise from HVAC systems, cooling towers, and power distribution infrastructure.30 Environmental monitoring near data center clusters has documented consistently elevated noise levels at residential distances — 22.3 dB at 2 miles and 28.0 dB at 200 feet, with high-frequency sound from power distribution infrastructure affecting residential areas.30 Chronic environmental noise exposure may be associated with hypertension, ischemic heart disease, sleep disruption, and annoyance in the broader epidemiologic literature.25,26
Example tools: Continuous dB(A) monitoring at property boundary and nearest residence
Evidence basis: Environmental monitoring (Ngata et al., 2025); noise-health literature (Hahad et al., 2023; de Souza et al., 2015).25,26,30
✓
Pathway 3E
Waste Heat & Urban Heat Island

Data centers emit waste heat with modeled thermal dispersion extending beyond facility boundaries; community health impacts have not been directly measured at the time of this review.

Data centers can generate waste heat that may extend beyond facility boundaries and potentially contribute to localized ambient temperature elevation in proximate areas. Modeling indicates thermal plume dispersion extending up to 150 meters under low wind conditions, with variation based on meteorological conditions and facility configuration.31 Considerations regarding heat-related cardiovascular and respiratory morbidity, emergency department utilization, and mortality in vulnerable populations are documented in the broader heat-health literature, though none directly measure heat from data centers.32,33
Example tools: NOAA surface temperature · USGS/NASA NDVI · CFD thermal modeling
Evidence basis: Thermal modeling (EOLIOS, 2025); heat-health literature (Liu et al., 2022; Singh et al., 2024). No data center-specific community health outcomes identified.31,32,33
✓
Pathway 3F
Light Pollution & Vegetation Loss

Data center development may be associated with increases in nighttime light intensity and reductions in vegetation cover near facility clusters; health effects in adjacent populations have not been directly measured at the time of this review.

Data center development and operation may be associated with increased nighttime light and vegetation loss near facility clusters Satellite monitoring near data center clusters has documented vegetation decline (NDVI reduction) and nighttime light intensity increases spatially associated with infrastructure expansion.20 Artificial light at night may be associated with sleep and circadian disruption in other exposure contexts.34 Green space loss may be associated with reduced physical activity, elevated heat exposure, and mental health outcomes in other infrastructure contexts.21,22
Example tools: VIIRS nighttime radiance · USGS/NASA NDVI satellite data
Evidence basis: Satellite monitoring (Chakraborty, 2024); analogous light and green space health literature (Blume et al., 2019; Browning et al., 2022).20,21,34
✓
Pathway 3G
Economic Burden & Energy Equity

Data center electricity demand growth may contribute to regional utility rate increases affecting household energy burden; downstream health effects have not been directly measured at the time of this review.

Increased electricity demand from data center growth may contribute to regional utility rate increases, with potential downstream implications for household energy burden and access to essential resources. Data center electricity demand growth may contribute to regional utility rate increases affecting households in the service area.18 Energy burden — defined as the proportion of household income spent on electricity — has been described in the broader literature in relation to social determinants of health including food insecurity, healthcare access constraints, and housing instability.35,36
Example tools: ACEEE LEAD tool · USDA ERS food security data · Local utility rate data
Evidence basis: Economic analyses and theoretical reviews (Jessel et al., 2019; Tuttle & Beatty, 2017); not measured in data center-specific contexts.35,36
✓
Pathway 3H
Power Quality & Grid Stability

Large data centers may introduce harmonic distortion and voltage variability into local distribution infrastructure; community health implications have not been directly measured at the time of this review.

Large-scale data center electrical draw may introduce disturbances into local distribution infrastructure, though community health implications are not well characterized in the current literature. Electrical infrastructure associated with large facilities may introduce harmonic distortion into local distribution grids, potentially affecting residential appliances and electrical systems in the service area. Voltage fluctuations and grid stability considerations have been discussed in grey literature related to communities near large facilities.
Example tools: Coordination with utility operators · Total harmonic distortion monitoring
Evidence basis: Technical grey literature; limited peer-reviewed characterization of community health implications.

Operational Phase: Considerations

Information that may be useful, if available: electricity consumption — direct and upstream (Scope 1/2) with generation source attribution; water withdrawals — direct, upstream, and supply-chain (Scope 1/2/3); backup generator operation hours and estimated NOx emissions; noise monitoring results; and power quality data where available. Identifying which community is most relevant to each pathway — fenceline vs. power plant-adjacent vs. corridor communities — is important for potential impact understanding.14

4
Phase 4
End-of-Life Considerations
Evidence framing: End-of-life impacts are primarily characterized through global electronic waste projections, material composition analyses, and analogous e-waste health literature, rather than facility-specific decommissioning studies.37 Exposures may occur at locations geographically distant from the original facility — including e-waste processing communities and transport corridor communities in different jurisdictions.
Click any pathway card to add it to your summary for further review.
✓
Pathway 4A
Electronic Waste — Heavy Metals

Decommissioned server hardware contains heavy metals including lead, mercury, and arsenic; community health outcomes attributable to data center-specific e-waste have not been directly measured at the time of this review.

Server hardware contains heavy metals that may generate exposure risks during processing, with health concerns documented in the broader e-waste literature though not yet characterized in data center-specific community studies. Electronic components contain heavy metals including lead, mercury, arsenic, and copper, which may generate inhalation and ingestion exposures in processing facility workers and nearby communities when handled through improperly controlled channels.38 Health concerns described in the broader e-waste literature include developmental toxicity, neurotoxicity, cancer risk, and immune dysfunction.38
Example tools: Certified e-waste processor documentation · State solid waste agency data
Evidence basis: E-waste environmental and health literature (Tansel, 2022; National Academies, 2022); no data center-specific community studies identified.8,38
✓
Pathway 4B
PFAS in Semiconductor Components

Semiconductor components contain PFAS compounds detected in environmental media near e-waste processing sites; community health outcomes attributable to data center semiconductor waste have not been directly measured at the time of this review.

PFAS-containing materials used in semiconductor manufacturing represent a potential exposure pathway during e-waste processing, with environmental detections documented near processing sites. PFAS use in semiconductor manufacturing accounts for approximately 10% of total PFAS-containing fluoropolymers used in computing.39 PFOS and PFOA have been detected in soil, water, blood sera, and rainwater at or near e-waste processing facilities.39 Health concerns described in the broader PFAS literature include endocrine disruption, cancer, immune dysfunction, and thyroid disruption.38
Example tools: State PFAS reporting data · EPA PFAS resources
Evidence basis: PFAS environmental monitoring literature (Elgamal et al., 2025; Tansel, 2022); e-waste site detections documented; data center-specific community studies absent.39
✓
Pathway 4C
Lithium-Ion Battery Systems

Lithium-ion UPS systems present fire and toxic release hazards during transport and end-of-life processing; health outcomes in affected populations have not been directly measured at the time of this review.

Lithium-ion UPS battery systems present acute safety risks during transport and end-of-life processing, relevant to workers and communities along transport corridors. Uninterruptible power supply battery systems present fire and explosion risks during transport and processing at end-of-life.42,24 Acute injury and toxic smoke inhalation risks may occur in connection with lithium-ion battery thermal events and may be relevant to transport corridor communities and processing facility workers.42 This pathway is most relevant in the context of decommissioning planning and transport route characterization.42,24
Example tools: Specialized battery transport certification documentation · Local emergency management coordination
Evidence basis: Technical grey literature and battery safety literature (NFPA, 2026).42
✓
Pathway 4D
Refrigerants & Heat Transfer Fluids

Cooling system decommissioning requires handling of refrigerants; community health outcomes have not been directly measured at the time of this review.

Cooling system decommissioning involves handling of refrigerants and heat transfer fluids whose hazard profiles vary by formulation. Handling of these materials involves variable risks including greenhouse gas emission potential, ozone depletion potential, and acute toxicity in some refrigerant categories.44 Safe decommissioning of cooling systems is addressed in industry standards and EPA Section 608 requirements for refrigerant recovery.44
Example tools: Refrigerant inventory disclosure · Certified refrigerant recovery documentation
Evidence basis: Technical literature; EPA Section 608 refrigerant recovery requirements; industry decommissioning standards.8,44
✓
Pathway 4E
Equipment Transport Emissions

End-of-life equipment transport generates diesel PM2.5 and NOx along designated corridors; health outcomes in corridor populations have not been directly measured at the time of this review.

End-of-life equipment transport may generate diesel PM2.5 and NOx exposures along corridors that are geographically distinct from the original facility location and may affect communities with no prior connection to the facility.43 Transport of electronic waste and decommissioned server equipment via diesel trucking contributes to emissions that may disproportionately affect communities along designated transport routes.43
Example tools: Geospatial local routing and zoning data
Evidence basis: McNeil et al. (2025). Impact of truck electrification on air pollution disparities in the United States. Nature Sustainability.43

End-of-Life: Considerations

Information that may be relevant, if available: a Decommissioning Plan well in advance of facility closure, documenting: equipment inventory and disposition protocols; PFAS-containing component disclosure; battery transport and disposal certification; and transport route mapping for corridor community identification. The geographic scope of potential end-of-life impacts may extend beyond the facility's original service area and may involve different jurisdictional authorities.

EH
Cross-Cutting Module
Environmental Health Surveillance

About These Indicators

The surveillance indicators below reflect commonly available measures that are available to most health departments.41 Primary data sources are the CDC Environmental Public Health Tracking Network (EPHTN) for air quality, heat, CVD, asthma, and temperature indicators (ephtracking.cdc.gov · Heat Risk Tool), and the Behavioral Risk Factor Surveillance System (BRFSS) for mental health indicators (cdc.gov/brfss).40 No data center-specific environmental health surveillance infrastructure currently exists; establishing baseline monitoring prior to development can provide valuable longitudinal reference data.

1
Pre-Development Baseline
Potential indicators for monitoring baseline conditions prior to Phase 1 (Pre-Siting) or Phase 2 (Construction).
DomainIndicatorPrimary Source
Ambient Air QualityPM2.5 annual mean; ozone 8-hr; NO2EPA AQS / AirNow CDC EPHTN
Respiratory HealthAsthma prevalence; respiratory ED visit rateCDC PLACES CDC EPHTN
Cardiovascular HealthHypertension prevalence; CVD mortality rateCDC PLACES; State vital statistics CDC EPHTN
Heat-Related IllnessHeat-related ED visits; heat mortalityState syndromic surveillance; CDC WONDER CDC EPHTN
Temperature / Air QualityMean summer temperature; days exceeding AQI thresholdsNOAA; EPA AQS CDC EPHTN
Mental HealthSelf-reported poor mental health days; depression prevalenceBRFSS
Water QualityDrinking water quality parameters; contaminant violationsState SDWIS; local utility reports
Economic BurdenEnergy burden index; food insecurity rateACEEE LEAD; USDA ERS
Noise BaselinedB(A) at nearest residential receptorsDirect monitoring
Nighttime Light / VegetationVIIRS nighttime radiance; NDVIUSGS/NASA satellite data
2
Construction Phase Monitoring (Acute)
Potential indicators for monitoring during Phase 2 (Construction).
IndicatorConsideration
PM2.5 at site perimeter and nearest residenceNAAQS 24-hr standard CDC EPHTN
Noise at nearest residential receptorCompare to pre-construction baseline and local ordinance
Diesel equipment emissions complianceDocument Tier 4 engine use
Respiratory syndromic surveillanceCompare to pre-construction baseline and control areas CDC EPHTN
Community health inquiriesTrack and document all health-related community contacts
3
Operational Phase Health Outcome Monitoring (Chronic)
Potential indicators for monitoring during Phase 3 (Operations).
Health OutcomeData Source
Asthma ED visits & hospitalizationsState hospital discharge; syndromic surveillance CDC EPHTN
Cardiovascular events (MI, CHF)State hospital discharge; vital statistics CDC EPHTN
Heat-related illnessSyndromic surveillance; vital statistics CDC EPHTN
Mental health outcomesBRFSS; Medicaid claims if available BRFSS
Economic health indicatorsSNAP enrollment; Medicaid; food bank utilization
Birth outcomes (low birthweight, preterm birth)State vital statistics
All-cause mortalityState vital statistics
4
End-of-Life Phase Surveillance Indicators
Potential indicators for monitoring during Phase 4 (End-of-Life) and decommissioning.
DomainIndicatorPrimary Source
Air Quality — Transport CorridorsPM2.5 and NOx associated with transport corridorEPA AQS · AirNow CDC EPHTN
PFAS Environmental MonitoringSoil, groundwater, and surface water PFAS screening near e-waste processing sites; drinking water PFAS parametersState environmental agency data · EPA PFAS resources · State SDWIS
Wastewater Quality TrackingEffluent monitoring at facilities receiving decommissioning waste streams; wastewater treatment compliance dataState NPDES permit data · EPA ECHO database
E-waste Processor Community HealthRespiratory and cardiovascular ED visits in communities near processing facilitiesState hospital discharge data CDC EPHTN
CS
Cross-Cutting
Key Considerations from the Literature

Dashboard Basis

This considerations summary synthesizes findings from Armstrong et al. (2026) and is organized consistent with the EPIS (Exploration, Preparation, Implementation, Sustainment) implementation science framework,9 the ASTHO/NACCHO Health in All Policies evaluation tool for state and local health departments,10 and the CDC 10 Essential Public Health Services.11 This dashboard was co-designed with the Pima County Health Department through iterative engagement with an environmental health working group. Activities on this dashboard reflect the health planning matrix and indicator profile approach recommended for incorporating health into cross-sectoral decision-making.10

Air-Cooled Systems: Key Themes

  • Air-cooled systems use minimal direct (Scope 1) water on-site.12
  • Higher electricity demand relative to water-cooled alternatives increases higher upstream electricity demand increases indirect air quality and water burdens at power generation facilities (Scope 2).12
  • Communities near or downwind of fossil fuel power plants serving the grid — which may be geographically distant from the data center — may bear a greater share of modeled air quality burden associated with higher upstream electricity demand (Scope 2).14,18
  • Upstream (Scope 2) air quality impacts vary depending on the carbon intensity of the electricity source..14

Water-Cooled (Evaporative) Systems: Key Themes

  • Evaporative cooling involves substantial direct (Scope 1) water withdrawal — potentially 2–9 million gallons daily at hyperscale facilities.5,6
  • Reduced electricity demand relative to air cooling may lower upstream power plant air (Scope 2) and water burdens at power generation facilities.12
  • Use of potable water for cooling, where applicable, may have implications for water system capacity in water-constrained regions.15

Community Considerations

  • Fenceline communities (adjacent to the facility) are in closest proximity to on-site (Scope 1) sources, including generator emissions, noise, waste heat, light, and land use change.
  • Power plant-adjacent communities may bear upstream power plant air (Scope 2) and water burdens — frequently geographically distant from the data center location.14,28
  • Transport corridor communities are in closer proximity to diesel emissions associated with construction and end-of-life decommissioning phases.
  • E-waste processing communities may be affected at decommissioning — potentially in different jurisdictions with separate governing authorities.

Evidence Limitations: Critical Context

  • No empirical epidemiologic studies directly measuring health outcomes near data centers have been identified in the scoping review.
  • All quantitative burden projections cited in this dashboard are model-derived estimates, not measured population health outcomes.
  • Mandatory environmental performance reporting does not currently exist; data availability varies substantially by operator and jurisdiction.8,16
  • Non-disclosure agreements have constrained assessment capacity in some published analyses.
  • This dashboard communicates plausible pathways informed by available literature; it does not establish causal relationships or provide regulatory thresholds.
EV
Reference
Evidence Base & Citations

Primary Source

Armstrong, NS, Cullen, T, Neatherlin, J & Exum, NG (2026). Direct and Indirect Public Health Impacts of Data Center Development: A Scoping Review. Johns Hopkins Bloomberg School of Public Health, Department of Environmental Health and Engineering. Prepared for Pima County Health Department. OSF Protocol Registration: OSF.io/v7s53.

Additional Methodological Frameworks

This dashboard incorporates principles from:

  • EPIS (Exploration, Preparation, Implementation, Sustainment) implementation science framework9
  • ASTHO/NACCHO Health in All Policies evaluation tool for state and local health departments10
  • CDC 10 Essential Public Health Services (Assessment, Policy Development, Assurance, and Equity dimensions)11
  • ATSDR Exposure Pathway Methodology. atsdr.cdc.gov — Overview of the Exposure Pathway Evaluation
  • EPA Interim Cumulative Impacts Framework (2024)23
  • NASEM Cumulative Impact Assessment Report (2025)23

Key Evidence Gap

The absence of empirical epidemiologic studies directly measuring health outcomes in communities near data centers is the defining limitation of the current evidence base. All quantitative health projections referenced throughout this dashboard derive from exposure modeling rather than measured population health outcomes. This dashboard communicates potential pathways supported by inferential literature, not established causal relationships.

Full Reference List

1.DataCenterMap.com. Accessed March 2026. https://www.datacentermap.com
2.Congressional Research Service. Data Centers: Background and Policy Considerations. CRS Report R48646. 2025. https://www.congress.gov/crs-product/R48646
3.Bogmans C, Ganpurev G, Gomez-Gonzalez P, Pescatori MG, Thube A. Power Hungry: How AI Will Drive Energy Demand. International Monetary Fund Working Paper. 2025.
4.Shankar S. Enhancing energy efficiency in AI-powered data centers. International Journal of High School Research. 2024.
5.Li P, Yang J, Islam MA, Ren S. Making AI less “thirsty.” Commun ACM. 2025;68(7):54–61.
6.Mytton D. Data centre water consumption. Npj Clean Water. 2021;4(1):11. doi:10.1038/s41545-021-00101-w
7.IBM. Data center classifications. 2026. https://www.ibm.com/think/topics/hyperscale-data-center
8.Armstrong NS, Cullen T, Neatherlin J, Exum NG. Direct and indirect public health impacts of data center development: A scoping review. Johns Hopkins Bloomberg School of Public Health, Dept. of Environmental Health and Engineering. 2026. OSF Protocol Registration: OSF.io/v7s53
9.Aarons GA, Hurlburt M, Horwitz SM. Advancing a conceptual model of evidence-based practice implementation in public service sectors. Adm Policy Ment Health. 2011;38:4–23. [EPIS Framework]
10.ASTHO/NACCHO. Health in All Policies Evaluation Tool for State and Local Health Departments. Accessed 2026. astho.org/HiAP-evaluation-tool
11.CDC. 10 Essential Public Health Services. Centers for Disease Control and Prevention. Accessed 2026. cdc.gov/10EPHS
12.Chen L, Wemhoff AP. Characterizing data center cooling system water stress in the United States. ASHRAE Transactions. 2022.
13.Western Resource Advocates. Data Center Impacts in the Western US: Policy Solutions for Water and Energy Use. 2025.
14.Guidi G, Dominici F, Gilmour J, et al. Environmental burden of United States data centers in the artificial intelligence era. arXiv. 2024. arxiv.org/abs/2411.09786
15.Li P, Yang J, Islam MA, Ren S. Making AI less “thirsty.” Commun ACM. 2025;68(7):54–61. [Scope 3 water accounting]
16.Crawford K. Generative AI’s environmental costs are soaring—and mostly secret. Nature. 2024;626.
17.Hajiesmaili M, Ren S, Sitaraman RK, Wierman A. Towards environmentally equitable AI. arXiv. 2024. arxiv.org/abs/2412.16539
18.Han Y, Wu Z, Li P, Wierman A, Ren S. The unpaid toll: Quantifying and addressing the public health impact of data centers. arXiv. 2025. doi:10.48550/arXiv.2412.06288
19.Virmani S. Investigating the environmental sustainability of data centers. American Journal of Student Research. doi:10.70251/hyjr2348.311018
20.Chakraborty S. Towards a comprehensive assessment of AI’s environmental impact. arXiv. 2024. arxiv.org/abs/2405.14004
21.Browning MHEM, Rigolon A, McAnirlin O, Yoon H. Where greenspace matters most: A systematic review of urbanicity, greenspace, and physical health. Landsc Urban Plan. 2022;217(104233).
22.Barton J, Rogerson M. The importance of greenspace for mental health. BJPsych Int. 2017;14(4):79–81.
23.National Academies of Sciences, Engineering, and Medicine. State of the Science and the Future of Cumulative Impact Assessment. 2025. nationalacademies.org/projects/DELS-BEST-23-04
24.Whitehead B, Andrews D, Shah A. The life cycle assessment of a UK data centre. Int J Life Cycle Assess. 2015;20(3):332–349.
25.Hahad O, Rajagopalan S, Lelieveld J, et al. Noise and air pollution as risk factors for hypertension: Part I—epidemiology. Hypertension. 2023;80(7):1375–1383.
26.Zaman M, Muslim M, Jehangir A. Environmental noise-induced cardiovascular, metabolic and mental health disorders: a brief review. Environ Sci Pollut Res Int. 2022;29(51):76485–76500.
27.Eslami E. Air quality and greenhouse gas emissions assessment of data centers in Texas. arXiv [physics.ao-ph]. 2025. arxiv.org/abs/2509.21312
28.Hennessy EM, de Chalendar JA, Benson SM, Azevedo IML. Distributional health impacts of electricity imports in the United States. Environ Res Lett. 2022;17(6):064011.
29.Kollar A, Grady C. The relationship between data centers and the climate is a systems challenge: a spatial analysis of United States data centers. Environ Res Commun. 2025;7(11):111005.
30.Ngata W, Bashir N, Westerlaken M, et al. The cloud next door: Investigating the environmental and socioeconomic strain of datacenters on local communities. arXiv [csDC]. 2025. arxiv.org/abs/2506.03367
31.EOLIOS Ingénierie. Urban heat island impact study for data centers. June 2025. eolios.eu
32.Liu J, Varghese BM, Hansen A, et al. Heat exposure and cardiovascular health outcomes: a systematic review and meta-analysis. Lancet Planet Health. 2022;6(6):e484–e495.
33.Singh N, Areal AT, Breitner S, et al. Heat and cardiovascular mortality: An epidemiological perspective. Circ Res. 2024;134(9):1098–1112.
34.Blume C, Garbazza C, Spitschan M. Effects of light on human circadian rhythms, sleep and mood. Somnologie (Berl). 2019;23(3):147–156.
35.Jessel S, Sawyer S, Hernández D. Energy, poverty, and health in climate change: A comprehensive review of an emerging literature. Front Public Health. 2019;7:357.
36.Tuttle C, Beatty TKM. The effects of energy price shocks on household food security in low-income households. 2017. doi:10.22004/AG.ECON.260484
37.Chouksey A, Rajan AK, et al. The green paradox: The climate, environmental, and sustainability implications of artificial intelligence. Global Environmental Change Advances. 2025. doi:10.1016/j.gecadv.2025.100029
38.Tansel B. PFAS use in electronic products and exposure risks during handling and processing of e-waste: A review. J Environ Manage. 2022;316(115291).
39.Elgamal M, Wei MA, Brooks GY, Hills D. PFASware: quantifying the environmental impact of per-and polyfluoroalkyl substances (PFAS) in computing systems. In: Design, Automation and Test in Europe. 2025.
40.CDC Environmental Public Health Tracking Network (EPHTN). Centers for Disease Control and Prevention. Accessed 2026. ephtracking.cdc.gov; CDC Behavioral Risk Factor Surveillance System (BRFSS). Accessed 2026. cdc.gov/brfss
41.Frist WH, Angell SY, Ebi KL, et al. Critical steps to address climate, health, and equity. Health Affairs. 2025;44(2):171–178. PMID: 39841948
42.National Fire Protection Association (NFPA). Lithium-ion batteries and data center safety. NFPA Journal. February 2026. nfpa.org/lithium-ion-data-center-safety
43.McNeil WH, Porzio J, Tong F, et al. Impact of truck electrification on air pollution disparities in the United States. Nature Sustainability. 2025. doi:10.1038/s41893-025-01515-x
44.U.S. Environmental Protection Agency. Section 608 of the Clean Air Act: Refrigerant Management. EPA. Accessed 2026. epa.gov/section608
45.Downey L, Van Willigen M. Environmental stressors: The mental health impacts of living near industrial activity. J Health Soc Behav. 2005;46(3):289–305.
46.Albrecht G, Sartore GM, Connor L, et al. Solastalgia: the distress caused by environmental change. Australasian Psychiatry. 2007;15(Suppl 1):S95–S98.