
Data Centers, Energy Use, and Local Backlash
Research overview
Executive summary
CENTRAL FINDING Approval should follow enforceable cost and impact evidence, not an announcement or a national average. · Evidence cutoff: November 19, 2026.
What the election result shows
Ballotpedia initially reported that the limiting side prevailed in 40 of 41 tracked November 3 contests. Its expanded tally counted 46 of 47. Both are counts of tracked measures, not national polling results. The limiting side sometimes won by defeating a pro-development proposal or by supporting an advisory question, not by enacting a ban. The updated tally is evidence of strong resistance where the question reached the ballot, not proof that voters everywhere oppose every data center. [1] [2] A separate September survey found local opposition among major partisan groups, with substantially different shares; it did not measure the effect of the issue on individual races. [9]
Scale makes the dispute consequential. The International Energy Agency estimates approximately 485 TWh of electricity was used by data centers worldwide in 2025 and projects roughly 950 TWh for 2030 in its updated central outlook. Those totals include non-AI workloads. Growth is concentrated in particular places, making the local peak and the capacity to serve it as important as national energy totals. [3]
What a defensible decision requires
Ask who pays if the facility opens late, uses less than promised, or closes early. Require separate estimates for on-site water, water embedded in electricity supply, local pollution and noise, and construction and land effects. A credible development agreement ties claims to meters, reporting, enforceable limits and a responsible authority. A credible moratorium has a defined scope, a public work plan and an expiration rule.
Where costs cannot be assigned or impacts cannot be bounded, delay or deny within lawful authority rather than assuming the community will be made whole. Where a project is demonstrably financeable, buildable and compatible, staged approval can preserve local benefits without locking the public into the entire speculative buildout.
How to read the evidence
SCOPE U.S. local decisions; global context only when the source boundary supports it.
Four evidence classes
An official canvass measures how ballots were counted in a specific contest. An editorial tracker measures selected, coded contests, and its historical snapshots can differ. An agency estimate or peer-reviewed model quantifies an explicitly defined population or scenario. A permit, tariff, or agreement establishes a specific constraint if it has legal effect; it is not proof that the promised outcome occurred. For every consequential claim, ask which class of evidence supports it.
This report uses official documents for rules and individual decisions where available, the tracker for the comparison it compiled, and model-based studies for estimated consumption and associated impacts. Captions label geography, year, unit and scope. Conceptual diagrams do not contain operational measurements. [2] [3] [5]
Reading order
Start with the ballot chapter for what voters actually decided. Read the power chapter to distinguish peak demand, energy and retail bills. Use the impacts and case chapters to test what is transferable to a specific project. The policy chapter turns those distinctions into a siting and ratepayer-protection checklist.
Treat forecasts as conditional. A national electricity projection does not establish spare capacity on a local feeder, water supply in an August drought, or the price on a family’s bill. A local case is often useful for its decision process, not as a forecast for another community. Official orders and certified results can change after the evidence cutoff; verify them before a hearing, investment or publication.
One project, several different decision makers
DECISION RIGHTS No single vote or contract determines every outcome.
Who controls what
A local land-use authority may choose a permitted use, impose lawful operating conditions, or reject a rezoning. A utility must determine whether it can serve the load and through which facilities. A state regulator may review rates, tariffs and the recovery of utility spending. Regional grid institutions may influence resource adequacy and regional transmission. Water, discharge and air permits can involve different authorities again. An approval in one venue is not a substitute for all the others.
A local official promising flat household rates may lack rate-setting power. Likewise, a utility contract cannot solve a community’s noise exposure or determine how residents use a local road. Readers should identify the entity with jurisdiction before asking it to act.
Maintain a decision ledger
Log the decision authority, docket or permit number, current instrument, proposed change, expiration, responsible decision maker and observable consequence. Record what remains contingent: a town may approve site use while a grid connection, financing, or final water plan remains unresolved. Publish nonconfidential information in a format residents can understand without relying on developer promotional materials.
At each phase, reconcile permits, utility milestones and construction. An approved building phase should not silently expand its power or water entitlement. A change in tenant, cooling design or generator operations can alter the effects even if the original address and project name stay the same. Define which changes trigger a fresh review.
What a large load number means
UNITS MW is power. MWh is energy. A booked request is neither delivered energy nor realized peak.
From design to delivery
Design capacity is not annual consumption. For an illustrative 100 MW facility with 75% annual load factor, average power is 75 MW and annual use is about 657,000 MWh, or 657 GWh, using 8,760 hours. This is arithmetic, not a prediction. If the 100 MW describes IT equipment rather than the whole facility, a separate facility-energy assumption is required to account for cooling and electrical overhead.
Peak demand is critical when the system is stressed. Annual MWh matters for energy purchases, fuel and emissions. Neither alone tells a planner how much reinforcement a specific location requires. The hourly load profile, ramp, interconnection configuration and capacity available at the time all matter. [4]
Reconcile the queue
A credible forecast distinguishes early inquiries, active studies, executed agreements, phased construction and energized load. A project shopping multiple jurisdictions can appear in more than one queue. A single leased phase can use less than its electrical design limit. A developer’s requested MW does not mean that amount will arrive as dependable load.
Model an upside, a downside and a timing delay. Identify which investments are reversible or reusable and which require an irreversible commitment. Stress test the downside before assigning cost to other customers. Do not label every data-center MWh as AI demand; data centers also serve storage, communications, ordinary enterprise applications and other digital workloads. [3] [4]
Costs, rates and bills are not interchangeable
AFFORDABILITY Compare costs with and without the project in the relevant utility territory.
Follow the cost chain
Physical infrastructure can include dedicated connection costs, shared planning upgrades and capacity or energy purchases. An accounting entry may be assigned to an industrial class, recovered under an approved tariff, covered by an up-front developer contribution or shared under a regulator-approved methodology. A wholesale cost increase does not become a household bill increase by itself. The commission order, class allocation, timing, customer usage and fixed charges determine the retail result.
An average statewide price is descriptive, not a causal estimate of what a proposed local project would do. A home using more kWh may see a higher bill even if the rate is flat. Compare the same household usage or use weather-normalized evidence when possible. [6]
Ask three counterfactual questions
What investment would the system need anyway? Which incremental spending is unavoidable if the project arrives? Who is liable for incremental costs if the load is delayed, ramps down or leaves? Consider a dedicated substation that can be assigned directly, versus a line that serves other demand and might be useful later. Their cost treatment need not be the same.
A large load can spread fixed costs when its revenues are sufficient, or shift risk when commitments are weak and the utility must build ahead. Neither outcome should be assumed. Require a transparent base case and explicit downside cases. Minimum bills, collateral and exit terms may mitigate those risks, but only if their design, enforceability and sizing reflect the exposure.
Water, emissions and noise need separate records
PLACE AND TIME A low annual average can hide a local peak.
Draw the boundary first
On-site withdrawal is not the same as consumptive use; some withdrawals return to a system, and often at a different location or quality. Water used to generate electricity is upstream of the facility. A global modeled supply-chain estimate does not measure a particular water supplier’s reserve in a drought. Cooling design should be evaluated against the actual basin, seasonal temperature, drought commitments, source and discharge pathways. [7] [8]
A site’s greenhouse-gas accounting is not the same as local exposure to generator emissions. Corporate clean-power procurement and regional dispatch address different questions. Specify whether the emissions estimate is modeled location-based attribution, a contractual market-based figure or measured on-site combustion.
Build a physical record residents can verify
Ask for monthly and peak water demand, annual and peak electric demand, generator operating category, construction hours, and a baseline sound survey. Set conditions for future measurements at defined receptors and record the methods. Explain impacts that extend beyond the host parcel or tax district, including roads, watersheds and regional grid upgrades.
Verification should not force the release of security-sensitive plans or private records. Public summaries can disclose the legally relevant commitments and aggregate observations. Sensitive supporting details can be handled by authorized reviewers where justified. Preserve complaint channels, escalation deadlines and clear accountability when a limit is exceeded.
Local approval is a choice among alternatives
OPTIONS Compare approve, condition, phase, pause or deny under applicable law.
Compare alternatives with the same metric
A moratorium can create time to write a rule, collect baseline evidence or resolve a capacity question. Its benefit declines if it has no work plan, a vague scope or no sunset. A siting ordinance can define an ongoing standard, but only if the municipality has authority and staff to enforce it. A project-specific agreement can produce detailed commitments yet creates inconsistency if important terms remain confidential or depend on discretionary enforcement.
A ban can make a boundary clear. It may also shift activity, jobs or infrastructure elsewhere and should not be proposed without checking authority, vested rights and legal constraints. Staged approval can limit irreversible commitments when the next phase depends on externally verified performance.
Use benefits as evidence, not promises
Separate temporary construction activity, permanent on-site employment, county tax base, utility contributions, and supplier spending. Look at who earns the gains and who pays for incremental roads, water systems, and emergency services. Gross tax figures are not net fiscal benefits. Broader economic benefits can be real even if the host county’s benefit is smaller than the state’s.
A community benefit agreement is useful when it identifies a payer, deadline, recipients, remedy and a reporting mechanism. A nonbinding donation or unsupported job projection cannot stand in for ratepayer or environmental conditions. Preserve both the option to decline and the ability to revisit promises when facts change.
A one-page decision test
RELEASE GATE Evidence before commitment. Conditions before expansion.
The gate
Before approval, identify five pieces of evidence: (1) a current project definition with phases; (2) a grid and ratepayer plan that assigns incremental cost under downside cases; (3) utility and water capacity stress tests; (4) a permit-ready site impact assessment with enforceable limits; and (5) a monitoring and exit plan with a named authority. A missing item is a review finding, not a cue to fill the gap with assumptions.
If the project is approvable in stages, define what has to be verified before each increment. If a commitment has to be made early, document its reversibility. When protections depend on a contract, confirm that the term survives ownership changes and covers the duration of the underlying investment.
The decision record
For each benefit, report the baseline, assumptions, distribution and evidence. For each cost, identify the party responsible if the project underperforms or exits. For each local impact, name the trigger, measurement method, reviewer and enforcement tool. Write down the strongest dissenting concern and what additional evidence would change the outcome.
No diagram replaces an enacted ordinance, a permit, a final tariff or a vote record. This report is educational public-source research, not legal, investment, engineering, or utility advice. It uses original charts and conceptual diagrams and does not conduct facility inspections.
Exhibit 01 · Decision rights across a project
The actors need not approve in this order. The point is to identify each independent authorization and avoid assuming one stands in for another.
48PEGASUS Research original conceptual exhibit. Not empirical data.
Exhibit 02 · What the community is deciding
Concerns belong in parallel, rather than being reduced to one number or a generalized label for public support.
48PEGASUS Research original conceptual exhibit. Not empirical data.
Browse: White Papers · Reports · Downloads
