Skip to content

Environmental and Social Impacts of Data Centers

Exploring how data center size, location, and infrastructure needs intersect with water resources, electricity systems, and community priorities.

Team: Carbon & Code
Hackathon: September 15–16, 2026
Status: Preliminary research framework and illustrative scenario comparison.

Project summary: We propose using public data to compare centralized and distributed data center development scenarios in South Dakota. This page distinguishes illustrative calculations from measured findings. Spatial analysis, site-specific validation, and appropriate community review remain necessary.

People

Name Affiliation Role
Emmanuel Akwasi Opoku Grambling State University
Tokata Stands Not provided Team member
Kiahna Standing Bear Not provided Team member

Team Norms and Decision Making

Proposed norms for team confirmation:

  • Separate evidence, assumptions, and interpretations.
  • Respect Tribal sovereignty, community knowledge, and information-sharing boundaries.
  • Document uncertainty and consider alternative explanations.

Proposed decision rule: Seek agreement on the research question, methods, and public claims. Record unresolved disagreements rather than presenting them as shared conclusions.

Our Question

How can public data help compare electricity demand and exposure to water stress for a hypothetical 500 MW data center development in South Dakota, configured as either one large facility or a distributed network of smaller facilities?

What would count as progress?

  • Identify relevant datasets and their limitations.
  • Present a transparent approach to estimating power demand.
  • Compare configurations with equal aggregate electrical capacity.
  • Define a reproducible mapping workflow.
  • Identify questions requiring additional evidence and community guidance.

Question-framing checkpoint: Community involvement in defining this question is not documented in the supplied materials. Any continuation should seek locally appropriate guidance rather than allowing available datasets alone to determine research priorities.

Why This Matters

This project explores questions about electricity supply, water availability, infrastructure costs, land use, and the distribution of potential benefits and burdens.

Our goal is not to assume that every data center has the same impact. It is to identify the evidence needed to evaluate a particular proposal.

Potential audiences include Tribal Nations, local residents, planners, water managers, utilities, researchers, and developers. These are potential audiences—not claims of partnership, consultation, or endorsement.

What We Tried to Build

The intended artifact is a screening-level comparison combining facility information, water-risk context, and geographic boundaries.

Proposed pathway: Data Investigator—evaluate evidence, coverage, and assumptions before building a more complex model.

Component Current Status
Research question and scope Drafted for team review
Size-to-power reference table Included; original technical source unverified
Equal-capacity scenario comparison Included as illustrative arithmetic
Facility and water-risk map Planned; completed output not supplied
Reproducible notebook Planned; completed notebook not supplied
Community review process Not documented

Data and Evidence

Source Geographic Scope Intended Use Important Limitation
IM3 Open Source Data Center Atlas United States Locate facilities and examine available footprints Location and footprint do not establish actual electricity or water use
WRI Aqueduct Global Screen locations for water-risk context Risk indicators are not facility consumption or contamination measurements
South Dakota county and contextual data South Dakota Support geographic summaries Specific datasets and versions remain to be selected
Team-supplied size-to-power table Unspecified Illustrate possible power-demand ranges Original source and measurement boundaries remain unverified
Project and utility documentation Site-specific Validate loads, cooling, water sources, and project status Documents have not yet been collected for this analysis

For each dataset, record its producer, source URL, version, reference period, access date, units, geographic resolution, license, and known limitations.

Public availability is not blanket permission. Accessible information is not automatically sufficient for interpretation or decision-making. Do not publish sensitive locations, culturally sensitive knowledge, personal information, or community-controlled data without appropriate authorization.

Estimated Power Demand by Facility Size

Building area may provide a preliminary proxy for power demand when its relationship to electrical load is documented and appropriate for the facilities being studied.

The following values come from the team-supplied reference table. They are illustrative, unverified ranges—not universal engineering requirements.

Building Area (sq. ft.) Illustrative Power Range (MW) Supplied Voltage (kV; Unverified)
50,000 5–10 35
100,000 10–20 69
250,000 25–75 69–138
500,000 50–150 138
1,000,000 100–300 138–230

Source: Team-supplied screenshot. The original technical publication and date were not provided.

How to Read the Table

  • MW measures power, not annual energy consumption.
  • MWh and GWh measure energy consumed over time.
  • kV measures voltage, not electricity consumption.
  • The source does not specify whether power means IT load, total facility demand, contracted capacity, or another rating.
  • The source does not define whether area means building footprint, total floor area, or IT equipment space.
  • The first two rows imply 100–200 W per square foot; the remaining rows imply 100–300 W per square foot. There is no single consistent density range.
  • Voltage values are retained for traceability, not design guidance. Confirm connection requirements with the relevant utility.
  • The table does not cover 2 MW or 500 MW facilities. The scenarios below use independent assumptions.

Calculation Framework

Estimated power (MW)
= applicable area (sq. ft.) × power density (W/sq. ft.) ÷ 1,000,000

Average facility power (MW)
= stated facility capacity (MW) × assumed average load fraction

Annual energy (MWh)
= average facility power (MW) × hours in the year

Illustrative example: A facility averaging 10 MW throughout a 365-day year would consume:

10 MW × 8,760 hours = 87,600 MWh = 87.6 GWh

This is arithmetic, not a measured facility result.

Power Usage Effectiveness

Power usage effectiveness, or PUE, relates total facility energy to IT equipment energy over the same measurement period:

PUE = total facility energy ÷ IT equipment energy

Use PUE when converting IT energy to total facility energy. Do not apply it again to an estimate that already includes facility overhead.

Reference: PUE and measurement boundaries.

Centralized Versus Distributed Development

For this comparison, assume all MW values represent total facility electrical capacity.

An equal-capacity comparison requires one 500 MW facility versus 250 facilities of 2 MW each.

Comparison Item Centralized Scenario Distributed Scenario
Number of facilities 1 250
Capacity per facility 500 MW 2 MW
Aggregate capacity 500 MW 500 MW
Assumed average load fraction 80% 80%
Aggregate average power 400 MW 400 MW
Annual electricity use 3,504,000 MWh 3,504,000 MWh
Annual electricity use in GWh 3,504 GWh 3,504 GWh

Calculations assume 8,760 hours per year. The 80% load fraction is illustrative, not observed.

Main Takeaway

Under identical aggregate capacity and load assumptions, both configurations have the same modeled annual electricity consumption.

Distribution changes where demand occurs. It does not automatically establish lower energy use, water demand, or environmental impact.

Equal electrical capacity also does not necessarily mean equal computing output. A stronger comparison would account for workloads, equipment, cooling, and operating conditions.

Named-Project Verification

The original draft referenced a proposed “Gemini” project. Its identity, location, capacity, and development status are not verified here.

The 500 MW case is therefore hypothetical, not a confirmed description of that development.

Water-Use Boundary

We do not estimate water use from MW alone.

A defensible estimate requires cooling-system information, water sources, operating conditions, and clearly defined water-use metrics.

Keep these categories separate:

  • Water withdrawal
  • Water consumption
  • Indirect water use associated with electricity generation

Methods and Tools

Proposed Workflow

  1. Inventory evidence. Record sources, versions, units, geographic coverage, and reuse conditions.
  2. Check facility records. Distinguish existing, proposed, and projected facilities; check for duplicate buildings and campuses.
  3. Validate area and power fields. Prefer documented facility values. Label proxy estimates and preserve low/high assumptions.
  4. Prepare spatial data. Confirm coordinate systems and valid geometries. Keep counties, watersheds, service areas, and jurisdictions distinct.
  5. Add water-risk context. Associate locations with the selected indicator and retain missing-data flags.
  6. Compare scenarios. Hold aggregate capacity constant and test sensitivity to load, density, and efficiency.
  7. Present findings and limits. Include sources, uncertainty, and accessible captions.
  8. Review before sharing. Check public suitability and seek appropriate local interpretation.

Possible tools: Python, pandas, GeoPandas, Jupyter Notebook, or desktop GIS software.

These are proposed tools, not a claim that the workflow has already been executed.

Open the team code folder

Link retained from the supplied template; repository contents were not validated during this revision.

Working Visual or Output

The tables on this page are the current visible artifacts.

Add maps or figures only when the corresponding analysis exists. Each caption should explain the source, period, geography, assumptions, and interpretation limits.

What We Made

  • Main artifact: A preliminary project brief and comparison framework.
  • Reference material: A clearly qualified size-to-power table.
  • Reusable methods: Formulas and a proposed spatial workflow.
  • Community-facing output: Questions for developers and relevant authorities.
  • Pending deliverables: Spatial analysis, a reproducible notebook, and site-specific validation.

What We Learned

These statements come from reviewing the supplied material and checking arithmetic—not from a completed spatial study.

Claim Level Supported Statement
What we observed The supplied power table lacks an original technical citation and clear measurement boundaries.
What we calculated Matching 500 MW with 2 MW facilities requires 250 facilities.
What we infer conditionally Equal aggregate capacity and equal load fractions produce equal modeled annual electricity use.
What we do not know Actual site loads, water use, cooling systems, grid constraints, and community priorities.
What we should not claim That this framework proves contamination, establishes local harm, or demonstrates that distributed development is inherently more sustainable.

What Did Not Work or Remains Incomplete

No failed technical experiments were documented in the supplied materials.

The main unresolved issues are:

  • Missing provenance for the power table
  • Undefined area and electrical-load measurements
  • No completed spatial analysis supplied
  • No validated site-specific water or utility data supplied
  • No documented community review process

These gaps should guide the next stage of work.

What Remains Uncertain

The current materials cannot establish actual facility consumption, water contamination, local grid adequacy, emissions, household cost effects, or which configuration has lower overall environmental impact.

A water-risk overlay can provide context, but it cannot by itself show that a data center caused water stress or contamination.

What Would Strengthen the Analysis?

  • Verified facility and utility records
  • Measured or defensibly modeled loads
  • Cooling-system and water-source documentation
  • Suitable geographic and temporal resolution
  • Sensitivity testing
  • Locally informed interpretation and appropriate review

Stewardship checkpoint: Who could be affected by this interpretation? Who is absent? Who should help frame and review further work? Naming potential reviewers does not imply consultation, consent, or endorsement.

What Is Next

Technical Priorities

  1. Select and document the IM3, Aqueduct, and South Dakota contextual datasets.
  2. Find the original power-table source or replace it with a documented estimation method.
  3. Build the spatial comparison and report missing data.
  4. Validate named developments using primary project and utility records.
  5. Test how results change under different assumptions.
  6. If adding international comparisons, align dates, capacity definitions, and project-status categories.

Stewardship and Collaboration Priorities

Seek locally appropriate guidance on the research question, interpretation, and public outputs.

Compile planning and governance examples for qualified review. This project does not establish which legal powers or zoning tools apply in a particular jurisdiction.

Questions for Developers and Relevant Authorities

  • Do the stated MW figures describe IT load, total demand, or contracted capacity?
  • What are expected average and peak electrical loads?
  • How will demand change as the project develops?
  • Which cooling system and water sources are proposed?
  • What are projected annual and peak-day water withdrawals and consumption?
  • How would drought affect operations and water availability?
  • Which grid upgrades are required, and who would pay?
  • What performance information will be publicly reported?
  • How will affected communities and relevant Tribal authorities shape the process?
  • What commitments address monitoring, complaints, operational changes, and closure?

Who Should Be Involved Next

Potential contributors include:

  • Tribal leaders and designated representatives
  • Residents and community organizations
  • Local planners
  • Water authorities and conservation districts
  • Utilities and grid operators
  • Independent technical reviewers

Their involvement should help shape the question and interpretation—not simply approve a finished model.

Roles, authority, consent, and information-sharing expectations should be established through appropriate processes.

Code, Data, Citation, and Reuse

Sources and Attribution

  1. IM3 Open Source Data Center Atlas
    Proposed source for facility locations and available footprints. Record the exact release and required citation.

  2. WRI Aqueduct
    Proposed source for water-risk context. Record the selected version, indicator, and methodology.

  3. Google Data Centers: PUE and Measurement Boundaries
    Reference for PUE accounting—not the source of the power table.

  4. Team-supplied screenshot
    Source of the illustrative area, MW, and voltage values. Original provenance remains unresolved.

  5. OLC Climate Resiliency and Digital Sovereignty Learning Lab
    Curriculum attribution retained from the original template.

  6. OASIS Team Repository
    Project link retained from the supplied document. Confirm and preserve the upstream template citation and license before redistribution.

Reproduction and Reuse Checklist

  • [ ] Record dataset versions, access dates, units, and preprocessing steps.
  • [ ] Preserve licenses and required attribution.
  • [ ] Include software versions and notebook execution instructions.
  • [ ] Report assumptions, measurement boundaries, and missing values.
  • [ ] Include sensitivity cases where appropriate.
  • [ ] Distinguish scenario calculations from measured results.
  • [ ] Publish only authorized, public-safe material.

This document is a preliminary screening and communication framework. It is not an engineering design, legal opinion, environmental impact assessment, or statement of community approval.