AI Infrastructure Tools

Democratic tools for governing the physical systems behind AI

AI does not run only in the cloud. It runs through land, buildings, power, water, cooling systems, transmission lines, zoning decisions, utility rules, tax incentives, and public trust.

This page tracks democratic tools for governing that infrastructure. The tools listed here are not automatically anti-growth. They are ways for public institutions, communities, workers, consumer advocates, and civic organizations to ask the right questions before AI infrastructure becomes locked in: where should it be built, who benefits, who pays, what burdens are created, and what obligations should attach to projects that require major public and environmental resources?

We will update this page as we find more examples. Each tool begins with one anchor example, then can grow over time as similar cases emerge.

1. Land-Use Review and Local Permitting

What it is and why it matters:
Land-use review is one of the most direct democratic tools for governing AI infrastructure. Data centers are physical projects. They require sites, permits, zoning approvals, utility hookups, roads, construction plans, noise management, and public process. Local planning commissions and city councils can therefore become a front door for AI governance, even when the broader technology debate feels national or global.

This matters in the Race because land-use review can slow the automatic conversion of local resources into private compute infrastructure. It gives residents and local officials a formal setting to ask whether a project fits the surrounding community, what public costs it creates, and whether the claimed economic or strategic benefits justify those costs.

Example we are tracking: Colorado Springs, Colorado: Project Taurus
Project Taurus is a proposed AI data center at a former Intel site near Garden of the Gods Road and Centennial Boulevard. Local residents have raised concerns about water use, power demand, noise, construction impacts, location, and whether strategic or military arguments are narrowing local accountability. Local coverage has described packed and heated community meetings, with the proposal now functioning as a test case for whether local land-use processes can meaningfully scrutinize AI infrastructure before it is approved. Additional local coverage: The Gazette and KRDO.

What to watch:
Whether local review remains a meaningful public process, or whether existing zoning, economic-development pressure, or national-security framing makes approval effectively automatic.

2. Temporary Moratoria and Policy Pauses

What it is and why it matters:
A moratorium is a temporary pause on new applications or approvals. It does not have to be a permanent ban. Used well, it gives a city or county time to develop standards before a wave of projects defines the rules by default.

This matters in the Race because AI infrastructure can move faster than local governance. A pause can be a civic catch-up tool: it creates time to study power demand, water use, ratepayer exposure, land-use compatibility, community impacts, and regional coordination before public institutions lose leverage.

Example we are tracking: Reno, Nevada: Temporary data-center moratorium
Reno became the first local government in Nevada to pause new data-center applications while it considers longer-term rules. The Nevada Independent described the pause as the biggest step a Nevada government had taken in response to growing data-center opposition. GovTech also covered Reno as the first local government in Nevada to take this step. Additional coverage: GovTech.

What to watch:
Whether the pause leads to durable standards, disclosure requirements, cost-allocation rules, and regional coordination — or whether it simply delays approvals without changing the underlying governance framework.

3. State Guardrails and Public-Interest Conditions

What it is and why it matters:
State guardrails are legal rules that allow data-center development but attach public-interest conditions. These may include water-use standards, reclaimed-water requirements, noise studies, limits on nondisclosure agreements, local zoning protections, and rules preventing utilities from shifting costs onto households and small businesses.

This matters in the Race because states can convert AI infrastructure from a private siting decision into a public-governance question. State policy can define minimum standards across jurisdictions, reduce secrecy, preserve local authority, and require large projects to internalize more of the costs they create.

Example we are tracking: Florida: State data-center guardrails
Florida enacted data-center guardrails that include ratepayer protections, local zoning authority, limits on nondisclosure agreements, water-resource review, and a framework for large-scale data-center permitting. The Florida House bill analysis describes provisions requiring the Public Service Commission to develop large-load tariff and service requirements, prohibiting certain public-sector nondisclosure agreements, creating a consumptive-use permitting framework for large-scale data centers, and requiring additional review for large-scale data centers near residential property or schools. Additional coverage: Florida Governor’s Office, Florida House bill analysis, and GovTech.

What to watch:
Whether state guardrails become real administrative capacity — tariffs, permit standards, water review, public records, and enforcement — or remain broad political statements with limited practical effect.

4. Phased Approval for Large Projects

What it is and why it matters:
Phased approval breaks a very large infrastructure project into stages instead of treating the full buildout as a single yes-or-no decision. It can preserve public leverage by requiring additional review as a project expands, especially when future power demand, water use, air-quality impacts, cooling systems, or community burdens remain uncertain.

This matters in the Race because AI infrastructure projects can be enormous, fast-moving, and technically complex. Phasing can prevent early approval from becoming a blank check for later expansion.

Example we are tracking: Utah: Stratos Project in Box Elder County
The Stratos Project is a proposed large-scale data and energy campus in western Box Elder County. Utah’s public FAQ states that creation of the project area was approved by the MIDA Board of Directors and Box Elder County Commission, but says full development is expected to occur in phases with further planning, infrastructure coordination, state permitting, and community engagement. Local and national coverage has focused on the scale of the proposal, energy and water questions, Great Salt Lake concerns, and public opposition. Additional coverage: Box Elder County, Deseret News, and Utah Public Radio.

What to watch:
Whether phased approval creates genuine decision points with enforceable conditions, or whether it becomes a procedural label for a project that is politically committed from the start.

5. Utility Cost Allocation and Ratepayer Protection

What it is and why it matters:
Utility cost allocation determines who pays for the grid upgrades, transmission projects, generation capacity, and service infrastructure required by large data centers. Ratepayer protection asks whether households and small businesses should subsidize infrastructure built primarily to serve large compute loads.

This matters in the Race because AI infrastructure can concentrate benefits while spreading costs. If public utility rules allow data-center-driven costs to be socialized across ordinary customers, then the buildout may function as a hidden transfer from the public to concentrated private infrastructure owners.

Example we are tracking: Maryland: Transmission cost-allocation challenge
The Maryland Office of People’s Counsel filed a complaint at the Federal Energy Regulatory Commission challenging PJM transmission cost-allocation rules. OPC argues that PJM’s rules could unfairly assign billions of dollars in data-center-driven transmission costs to Maryland customers. This reframes AI infrastructure as a ratepayer-protection and utility-governance issue: who should pay for the grid buildout required by large compute loads? Additional coverage: American Public Power Association, Reuters, and OPC complaint PDF.

What to watch:
Whether utility regulators and federal energy regulators require cost causation — assigning costs to the customers or zones that create them — or continue broad cost spreading that protects data-center economics while raising public bills.

6. Local Restrictions, Exclusions, or Bans

What it is and why it matters:
Local restrictions and bans are the strongest form of local infrastructure control. Instead of merely imposing conditions, a community may decide that data centers do not belong in a particular redevelopment area, zoning district, or local plan.

This matters in the Race because dispersion does not only mean distributing AI infrastructure everywhere. It can also mean preserving local democratic authority to say no, especially where projects would impose concentrated burdens without sufficient local benefit.

Example we are tracking: New Brunswick, New Jersey: Removing data centers from a redevelopment plan
In New Brunswick, the city council voted to remove data centers from a redevelopment plan after community opposition. Reporting described residents and environmental advocates raising concerns about energy use, water use, pollution, and impacts on residential communities. The case is useful because it shows local exclusion not as an abstract anti-technology position, but as a concrete land-use decision about what kinds of infrastructure a community wants to permit in a specific place.

What to watch:
Whether local bans remain isolated reactions, or whether they become a more common tool for communities that conclude data-center development is incompatible with local plans, environmental limits, or public priorities.