Putting Data Centers in Context: Promise, Burden, and the Case for Sound Governance
Try the Interactive Explorer To accept the example data and see the results, click the green "Evaluate" button in the lower right-hand corner of either:
- The Example Policies data (TOML)
- The Example Sites data (CSV)
Displayed results are illustrative. Try the interactive explorer for current examples.
This project provides a structured framework for exploring data center siting and governance tradeoffs under explicit assumptions and constraints. It:
- frames the issue as one of infrastructure governance
- recognizes the distribution of costs vs benefits
- emphasizes integration with broader systems rather than isolated analysis
The goal is to make tradeoffs visible and inspectable across multiple dimensions.
The contribution of this project is the framework for structured exploration, not the specific values used in any given evaluation.
- Constraints, thresholds, and weights are configurable
- Assumptions are explicit and inspectable
- Results are comparative and assumption-dependent
This project does not determine outcomes or recommend decisions. It provides a way to examine how different assumptions and constraints shape outcomes.
Working files are found in these areas:
- data/ - source inputs and scenario configuration
- docs/ - narrative, assumptions, and analysis
- src/ - implementation
- Loads candidate sites from CSV and policy constraints from TOML
- Evaluates hard-constraint admissibility for each site (PASS / FAIL)
- Exports results as JSON for the web Explorer
- Interactive web Explorer for non-technical users
Tax revenue is potentially significant and is set during negotiations. In West Des Moines, Iowa, Microsoft's data centers are projected to generate over $2 billion in tax revenues over the agreement period (Brookings). Loudoun County, Virginia (the largest data center market in the world) now receives an estimated $890 million annually in data center tax revenue, nearly matching its entire operating budget, and has lowered its residential real estate tax rate incrementally as a result.¹
Data center demand can accelerate grid upgrades that benefit all ratepayers, not just the facility (Brookings).
Some operators have built regional fiber networks as part of development agreements, enabling businesses, students, and telemedicine across rural areas that would otherwise lack connectivity (Brookings).
Microsoft partnered with Gateway Technical College in Wisconsin to launch a Datacenter Academy training more than 1,000 students in five years, and partnered with the University of Wisconsin-Madison on AI-driven research (Brookings).
¹ Loudoun County figures: Cardinal News
The Memphis case presented here can be documented at greater length as it is the most fully reported U.S. case study currently available.
The xAI Colossus facility in South Memphis began operations in summer 2024, powered by portable natural gas turbines installed without air permits. Researchers at the University of Tennessee, Knoxville, analyzing satellite data from NASA and ESA, found that peak nitrogen dioxide concentrations in areas immediately surrounding the facility increased by 79% compared to pre-facility levels. Ozone readings in the Memphis metropolitan area have exceeded federal limits for at least two years, and the Southern Environmental Law Center (SELC) petitioned the EPA to formally recognize the area's failure to meet national air quality standards. The American Lung Association gave both Shelby County, TN, and DeSoto County, MS an "F" for ozone pollution.
Sources:
Note on contested evidence: Researchers at the University of Memphis, using air dispersion modeling and an independent two-day monitoring campaign, concluded that xAI's turbines had not measurably degraded ambient air quality, though they noted the increases would compound an already above-limit baseline for fine particulate matter. The two analyses used different methods and measurement periods; both are cited here (The Conversation).
In 2024, state officials allowed xAI to operate turbines without a permit by classifying them as "temporary" and "mobile," with no required tracking of toxic releases. After the NAACP and SELC issued a formal notice of intent to sue, xAI removed 20 turbines from the Colossus 1 site and obtained permits for the remaining 15. The company then repeated the same approach at Colossus 2, despite public opposition, with xAI publicly stating it planned to "copy and paste" the same strategy. As of April 2026, the NAACP, represented by SELC and Earthjustice, has filed a federal Clean Air Act lawsuit.
Sources:
The Colossus facility borders Boxtown, a community that is 90% Black with a median income of approximately $36,000. Shelby County already led Tennessee in asthma hospitalizations prior to the facility opening. Boxtown was annexed from the City of Memphis in the 1960s as part of an "urban renewal" program and has since become home to multiple industrial facilities, including a coal oil refinery, chemical plants, and large-scale food processing operations. Community members testified at public hearings, with one resident stating through tears: "I can't breathe at home, it smells like gas outside."
Sources:
Colossus cooling systems are projected to eventually draw upward of 5 million gallons per day from the Memphis Sand Aquifer, a regional drinking water source. The aquifer is overlain by unlined coal ash ponds containing arsenic, raising concerns among community groups that increased pumping could draw contaminated water downward into the drinking water supply. The nonprofit Protect Our Aquifer notes that more water is currently being withdrawn from the Memphis Sand Aquifer than is being naturally replenished. xAI has committed to an $80M water recycling facility using treated wastewater by end of 2026, but community advocates have raised concerns about oversight given the company's regulatory track record.
Sources:
Memphis estimates receiving $13 million in tax revenue from xAI in its first year, with only $3 million allocated to zip codes within a five-mile radius of the facility: the area bearing the direct environmental costs. State Representative Justin Pearson, who represents South Memphis, has stated publicly that the billions in company valuation do not translate to comparable local investment (NCRC).
- International Energy Agency (IEA) Energy and AI (April 2025): https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai
- Global: ~415 TWh in 2024 (1.5% of global electricity); projected to double to ~945 TWh by 2030 (~3%)
- U.S.: ~45% of global total in 2024; ~4% of all U.S. electricity
- Nearly half of U.S. capacity concentrated in five regional clusters (Northern Virginia, Dallas, Silicon Valley, Phoenix, Chicago area)
- U.S. data centers projected to consume more electricity by 2030 than all energy-intensive manufacturing combined
- AI is primary driver; accelerated servers growing ~30% annually
- U.S. Energy Information Administration (EIA): https://www.eia.gov/
- U.S. Department of Energy (DOE): https://www.energy.gov/
- Lawrence Berkeley National Laboratory: https://datacenters.lbl.gov/
- Electric Power Research Institute (EPRI): https://www.epri.com/
- U.S. Geological Survey (USGS): https://www.usgs.gov/
- USGS Water Use in the United States: https://www.usgs.gov/mission-areas/water-resources/science/water-use-united-states
- World Resources Institute (WRI): https://www.wri.org/
- National Ground Water Association (NGWA): https://www.ngwa.org - publishes research on data center groundwater demand and aquifer impacts
- National Conference of State Legislatures (NCSL): https://www.ncsl.org/
- Federal Energy Regulatory Commission (FERC): https://www.ferc.gov/
- EPA Environmental Justice Screening Tool (EJScreen, archived): https://19january2021snapshot.epa.gov/ejscreen/index.html - environmental justice screening and mapping tool; removed from EPA website February 2025;
- EPA EJScreen v2.3 (reconstructed): https://pedp-ejscreen.azurewebsites.net/ - reconstructed and hosted by the Public Environmental Data Partners (PEDP)
- Public Environmental Data Partners (PEDP): https://screening-tools.com - volunteer coalition preserving and reconstructing federal environmental datasets
- CDC/ATSDR Social Vulnerability Index (SVI): https://www.atsdr.cdc.gov/placeandhealth/svi/index.html - census-tract level vulnerability data for identifying overburdened communities;
- EPA AirNow: https://www.airnow.gov - real-time and historical air quality index data by location
- Resources for the Future (RFF): https://www.rff.org - economics research on energy, environment, and infrastructure tradeoffs
- Bipartisan Policy Center: https://bipartisanpolicy.org - energy and infrastructure policy analysis
- BloombergNEF: https://about.bnef.com - data center energy demand forecasting from investment research perspective
- National Academies of Sciences, Engineering, and Medicine: https://www.nationalacademies.org - consensus scientific reports on energy infrastructure and environmental impacts
- Uptime Institute: https://uptimeinstitute.com/
- Green Grid: https://www.thegreengrid.org/
Show command reference
After you get a copy of this repo in your own GitHub account,
open a machine terminal in your Repos folder:
# Replace username with YOUR GitHub username.
git clone <https://github.com/username/decision_explorer_data_centers
cd decision_explorer_data_centers
code .# Set Up the Environment
uv self update
uv python pin 3.14
uv sync --extra dev --extra docs --upgrade
uvx pre-commit install
# Local format + lint
uv run ruff format --check .
uv run ruff check .
# Pre-commit (enforce repo rules)
git add -A
uvx pre-commit run --all-files
# repeat if changes were made
git add -A
uvx pre-commit run --all-files
# Static + security + dependency checks
uv run validate-pyproject pyproject.toml
uv run deptry .
uv run bandit -c pyproject.toml -r src
# Tests (after static checks pass)
uv run pytest --cov=src --cov-report=term-missing
uv run python -m decision_explorer_data_centers.cli --candidates data/raw/example_candidates.csv --policy data/raw/example_policy.toml --output-json docs/data/results.json
# Docs build (after everything passes)
uv run zensical build
# Commit and push
git add -A
git commit -m "update"
git push -u origin main