Boston Data Center Ban Proposal Turns the AI Boom Into a Land-Use Fight

Boston City Councilor Miniard Culpepper has proposed a Boston data center ban that would prohibit new facilities across every city zoning district. The measure arrived despite Culpepper saying he knows of no pending data center proposal in Boston. That timing is the point. City officials are considering whether to close the door before an AI infrastructure developer starts knocking.

The proposal would cover facilities built primarily for artificial intelligence, cloud computing, cryptocurrency, data storage, processing, transmission, and equipment colocation. Existing data centers could remain open, but they could not expand under the proposed rules. Ordinary server rooms used by businesses, telecommunications providers, and research institutions would remain permissible.

That distinction turns a local zoning amendment into a wider test. Massachusetts is trying to attract AI investment while containing its demands on electricity, water, land, and nearby communities. Boston’s proposal rejects case-by-case regulation as the default. It starts from prohibition and makes any future developer argue for an exception.

What the Boston Data Center Ban Would Actually Do

The proposed amendment would make data centers a prohibited use throughout Boston, without shutting down ordinary computing operations or existing facilities.

Culpepper filed the council order on September 14, 2026. The official zoning amendment was placed on the City Council agenda for September 16.

The text would add a data center definition to Boston’s zoning code. It would then classify that use as prohibited in every district and subdistrict. That citywide approach is broader than assigning data centers to industrial areas or requiring additional environmental reviews.

The definition focuses on a facility’s primary function. It reaches buildings dedicated to large-scale digital storage, processing, transmission, or computing. It also covers cryptocurrency mining, AI computation, cloud platforms, and colocation facilities serving multiple customers.

A hospital running servers for medical records would not automatically become a data center under that standard. Neither would a university laboratory with computing equipment needed for its research. Telecommunications infrastructure and equipment supporting a company’s main business would also remain outside the intended prohibition.

Those exclusions matter because nearly every modern organization operates some kind of server infrastructure. A definition based only on equipment or power consumption could capture offices, hospitals, laboratories, and network facilities unintentionally. The proposal instead targets properties whose central business is operating computing infrastructure.

Boston already has a large commercial data center at One Summer Street in Downtown Crossing. Markley Group operates the facility, which occupies approximately 920,000 square feet. WBUR reported that it opened in 1998 and uses 30 megawatts of electricity, roughly equivalent to the consumption attributed to 30,000 households.

That facility would be grandfathered, meaning it could continue operating because it predates the proposed zoning restriction. However, the amendment would prevent expansion that requires new zoning approval. Other existing facilities would receive the same treatment.

Grandfathering avoids an immediate fight over closing infrastructure that supports businesses, networks, and digital services. It also reveals the proposal’s forward-looking nature. Boston is not attempting to erase its existing computing base. It is deciding whether scarce city land should host another purpose-built computing campus.

The measure is not yet law. The council filing begins a process that can include committee review, public testimony, and revisions. The Boston Zoning Commission must approve the zoning change before it takes effect.

A developer might also seek a zoning variance, depending on how the final amendment is written and applied. A variance is permission to depart from zoning rules based on a property-specific case. That route would move a project from routine approval into a public, discretionary process.

This difference is central to the proposal. Boston would no longer wait for a developer to reveal a project before asking whether the use fits the city. It would establish incompatibility as the baseline, then require the developer to overcome it.

Culpepper summarized the case during the council meeting. He cited Boston’s density, limited developable land, and demands on existing infrastructure. He told WBUR that the city was trying to get ahead of projects that might still be taking shape.

The absence of an announced project weakens any claim that Boston faces an immediate construction threat. Yet it also gives officials more room to write rules without negotiating around a specific investment, site, or deadline. The policy question can be considered before sunk costs reshape the debate.

That is what changed. Data center development in Boston would cease to be an undefined use that might fit within a broader zoning category. It would become an explicitly prohibited land use, subject to public scrutiny before a developer could proceed.

Why Boston Is Moving Before a Developer Arrives

Boston’s proposal treats urban land and infrastructure capacity as resources that should be allocated before an AI company creates political pressure around a specific site.

Large data centers are difficult urban neighbors because their economic value does not depend on proximity to most city customers. Computing workloads can often run wherever developers can secure power, fiber connections, cooling capacity, and suitable land. A dense city therefore has to compare a data center against uses that need an urban location.

That comparison includes housing, laboratories, offices, light industry, public facilities, and commercial space. Each use competes for land, utility capacity, construction labor, and political attention. Boston’s housing constraints make that opportunity cost especially visible, even when a proposed site is not zoned for housing.

Data centers also concentrate infrastructure demand. A major facility can create a large, steady electrical load rather than thousands of smaller loads distributed across many properties. Utilities may need new substations, transmission upgrades, or generation resources to serve it.

The concern is not that every data center automatically raises household bills. Electricity rates result from several interacting factors, including generation, transmission, distribution, regulation, and procurement. The policy dispute concerns who pays when one large customer triggers expensive additions to the system.

Massachusetts now says developers should carry those incremental costs. Governor Maura Healey’s September 8 executive order requires certain projects to meet energy, environmental, and community conditions before state agencies advance their permits.

The order applies its central permitting requirements to data centers with peak electricity demand above 25 megawatts. Developers must conform to the state framework and submit a community benefits agreement developed with relevant stakeholders.

A community benefits agreement records commitments that a developer makes to residents or local institutions. Those commitments can cover workforce opportunities, environmental protections, infrastructure, monitoring, or financial support. The state order puts that process before key permit approvals.

The executive order also directs agencies to create an alternative compliance payment mechanism. A facility that does not procure enough new clean electricity for its annual consumption would make payments benefiting ratepayers.

Healey had already paused applications for Massachusetts’ data center sales and use tax exemption in June. Her administration says projects should bring new clean energy, protect water resources, avoid shifting costs to other customers, and engage communities transparently.

Boston’s proposal builds on that local-control principle but reaches a more restrictive conclusion. The state framework says a qualifying project can proceed when it earns local approval and meets guardrails. Culpepper’s amendment says Boston’s default answer should be no.

This conflict is not simply government versus technology companies. It is regulation versus prohibition as competing responses to the same infrastructure pressures.

A regulated path assumes developers can reduce or compensate for harm. Officials might require clean power, recycled water, noise limits, emissions controls, workforce agreements, and infrastructure payments. Each project would then be judged against enforceable standards.

A prohibition starts from Boston’s physical constraints. Even a comparatively efficient facility still occupies land and consumes power. Its cooling equipment and emergency systems still impose operational demands. A developer can reduce those demands, but it cannot remove the project’s basic footprint.

AI growth strengthens the argument for acting early. Lawrence Berkeley National Laboratory estimated that American data centers consumed 176 terawatt-hours of electricity in 2023, representing 4.4 percent of total U.S. consumption. Its 2025 update estimates they could reach 11.8 percent by 2030, with a scenario range of 9.5 to 15.3 percent.

Those are national projections, not a forecast for Boston. Regional grid conditions, facility types, climate, and efficiency vary substantially. However, the electricity forecast explains why cities are no longer treating server buildings as ordinary industrial development.

Modern AI systems intensify the issue because accelerated servers use specialized processors at high utilization. The resulting heat must be removed continuously. Operators can use mechanical refrigeration, evaporative cooling, liquid cooling, or combinations of those systems.

Each approach shifts the resource balance. Refrigeration can increase electricity demand, while evaporative cooling can increase water consumption. Liquid cooling can improve heat transfer but still needs systems that move and ultimately reject the heat.

Boston officials are therefore debating more than one hypothetical building. They are deciding whether the city wants to reserve a path for an infrastructure category expected to grow rapidly nationwide. Waiting for an application would move that decision into a developer’s schedule.

The Real Contest Is Local Control Versus Regulated Growth

The Boston data center ban makes local discretion the gatekeeper, while Massachusetts is building a framework intended to permit responsible projects.

Healey’s state policy tries to hold two positions together. Massachusetts wants AI research, investment, and related employment. It also wants to prevent developers from transferring environmental and grid costs to the public.

The state has invested in that technology agenda. A 2024 economic development law allocated $100 million to the Massachusetts AI Hub. Universities, hospitals, biotechnology companies, and software businesses all depend on growing computing capacity.

Not every workload needs to run inside Boston. Cloud services already distribute computing across regional and national facilities. Universities and companies can also use dedicated centers outside the city. Still, location affects network latency, resilience, data movement, and access to specialized infrastructure.

That creates the strongest argument against a permanent citywide prohibition. Boston could exclude facilities that support research, medical computing, telecommunications, or emerging businesses even when a project offers measurable public benefits.

The proposal attempts to address that concern through exclusions for computing that supports another principal use. A university could run research servers, for example, without turning its campus into a prohibited data center. The hard cases would involve facilities serving many institutions or customers.

Holyoke shows why those cases cannot be judged by the word “data center” alone. The Massachusetts Green High Performance Computing Center supports research by universities including Harvard, MIT, Boston University, Northeastern, UMass, and Yale.

Holyoke enacted a near-total ban on new data centers but preserved that existing nonprofit facility. Its center occupies a remediated former industrial site and relies heavily on the city’s low-carbon electricity supply.

The facility also recirculates cooling water and uses outside air for cooling during much of the year. It consumes between 20,000 and 30,000 gallons of water daily and is permitted to draw up to 12 megawatts, according to a detailed Holyoke case study.

Those figures still represent meaningful resource consumption. Yet local leaders describe the facility as a good neighbor because it limits nuisance impacts, supports research, provides community programs, and contributes payments to the city.

Holyoke’s experience supports both sides of Boston’s argument. It demonstrates that a well-sited data center can deliver public value with fewer local conflicts. It also shows that those outcomes depend on location, ownership, design, and community relationships rather than generic industry promises.

Lowell provides the opposing reference point. Residents near Markley Group’s facility have complained about noise, diesel generators, emissions, and water use. The city approved a one-year moratorium on new construction and expansion as officials reconsider zoning rules.

Markley has defended its operations. The company has said its generators run during actual outages and short testing periods rather than continuously. It also told the Associated Press that the facility uses about 118,000 gallons of water daily during peak summer conditions.

Residents and environmental advocates dispute whether existing protections are adequate. A lawsuit challenged Massachusetts regulators’ approval of an air-quality plan involving additional diesel generators. A judge temporarily blocked the installation of four generators while allowing four already installed units to operate.

These examples place pressure on developers to offer more than broad sustainability commitments. A Boston applicant would need site-specific evidence showing its electricity supply, water requirements, backup generation, noise profile, emissions, and community value.

It would also need to explain why the workload belongs inside Boston. Cheap land is not available, and the urban grid has competing demands. A project justified mainly by access to Massachusetts customers might be built elsewhere in the region.

The Boston data center ban changes that negotiation. Under permissive zoning, opponents must prove that a proposed facility creates unacceptable harm. Under prohibition, the developer must prove that exceptional circumstances justify approval.

That shift can improve public leverage. Developers would need community engagement before treating approval as an expected outcome. Officials could examine infrastructure costs without being told that delays threaten a nearly completed transaction.

It can also reduce predictability. Companies prefer clear technical standards because they can design around them. A discretionary exception depends on political judgment, which can change across sites and administrations.

Massachusetts’ framework offers one route to predictable approval. It sets expectations around clean energy, cost allocation, environmental protection, and community agreements. Boston’s approach would place a local threshold above those state requirements.

The central policy choice is therefore not whether data centers create benefits or costs. Both are real. The question is whether Boston believes enforceable conditions can make a new facility worthwhile within its unusually constrained urban setting.

A Blanket Prohibition Still Faces Hard Questions

The case for early action is credible, but calling the amendment a ban can obscure unresolved questions about definitions, exceptions, and regional displacement.

The first question is scope. Data centers range from compact network facilities to hyperscale campuses. Their power demand, water use, emissions, staffing, and physical footprint differ enormously.

Boston’s proposal uses a functional definition rather than a single size threshold. That approach can stop developers from dividing a project into smaller components. It can also create ambiguity when a facility supports telecommunications, research, cloud services, and outside customers simultaneously.

Colocation is a particularly difficult category. A colocation provider leases secured space, power, cooling, and network access to many customers. These facilities can support hospitals, financial institutions, government agencies, startups, and network operators without controlling the workloads inside.

Treating every colocation site like an AI training campus would ignore those differences. Exempting colocation broadly could create a loophole that allows an AI-heavy facility to avoid the ban by leasing racks to separate customers.

The amendment’s review process must resolve that tension. Regulators need criteria based on primary use, scale, electrical demand, and operational characteristics. Otherwise, businesses could face uncertainty about whether an ordinary infrastructure upgrade counts as prohibited expansion.

The second question is displacement. Blocking data centers inside Boston does not reduce demand for computing. It moves the facilities and their impacts to another community.

That outcome might be reasonable when land and grid capacity are better suited elsewhere. It becomes less defensible if cities with political influence export infrastructure to lower-income communities while retaining the economic benefits of the digital services it supports.

Massachusetts’ environmental justice rules are intended to reduce that risk. Yet local vetoes can produce a regional pattern that no single municipality plans. Communities compete for tax revenue while sharing electricity markets, watersheds, air pollution, and transmission systems.

A citywide Boston restriction should therefore be evaluated alongside regional capacity planning. Officials need to know where displaced demand would go, what infrastructure would serve it, and which residents would bear its impacts.

The third question concerns public benefits. Data centers usually create substantial construction work, but their permanent staffing can be modest relative to their size and utility consumption. They can also provide property-tax revenue from land that might otherwise remain underused.

Boston must compare those benefits with realistic alternatives for each site. A blanket rule avoids speculative projections, but it also prevents planners from considering a highly beneficial project through ordinary zoning.

The variance process may become the pressure valve. If exceptions are possible, the city can require a developer to establish unusual hardship and public value. However, frequent variances would undermine the clarity promised by a citywide ban.

The fourth question is whether Boston is solving an immediate local problem or signaling a political position. Culpepper said he was unaware of any new proposal in the city. Boston’s only cited large commercial facility has operated for decades.

That does not make the policy unnecessary. Zoning is supposed to guide future land use. Still, the lack of an active project means officials have limited local data about likely sites, grid connections, water demands, or economic offers.

A moratorium could provide time to collect that information while preserving future options. Several Massachusetts communities have chosen temporary pauses rather than permanent bans. Boston would instead establish prohibition and revise it later if circumstances change.

The fifth question is technological change. More efficient chips, improved cooling, heat reuse, and flexible power consumption can lower a facility’s impact. Conversely, efficiency gains can encourage more computing, leaving total demand higher.

A zoning code cannot assume either outcome. It needs rules that remain useful as equipment changes. Facility-level electricity demand and environmental effects are more durable regulatory targets than a specific computing technology.

Public opinion gives elected officials a strong incentive to choose caution. National polling reported in September found broad opposition to data centers near respondents’ homes. Local conflicts have emerged across political lines, often centered on utility costs and environmental impacts rather than opposition to computing itself.

However, public concern does not settle the technical questions. A responsible decision must distinguish verified local effects from national projections and worst-case examples.

Lowell’s experience is relevant because it involves the same operator as Boston’s major existing facility. Yet Boston’s density, water systems, electrical infrastructure, building types, and neighborhood exposure are different. Officials should not assume identical outcomes.

The strongest version of the ban is therefore procedural, not symbolic. It should close unclear zoning pathways, create a public burden of proof, and preserve narrow exceptions with transparent standards.

The weakest version would merely declare opposition while leaving broad variance routes and ambiguous definitions. That could produce prolonged disputes without giving communities, developers, or utility planners clear expectations.

What Happens Next Will Decide Whether the Ban Has Teeth

Three decisions will show whether Boston is establishing a durable land-use policy or opening a negotiation over exceptions.

The first signal is the City Council process. Committee hearings should reveal whether other councilors support permanent prohibition, a temporary moratorium, or conditional zoning standards.

Ed Flynn previously filed an order calling for a hearing on an AI data center moratorium. Nine other councilors joined that filing, showing substantial interest in stronger oversight. Support for studying the issue does not automatically translate into support for Culpepper’s citywide prohibition.

Testimony from utilities, environmental groups, researchers, businesses, labor organizations, and residents will shape the final definition. The most important details include how the amendment handles colocation, telecommunications facilities, research computing, and expansions at existing sites.

A council decision to preserve the citywide prohibition with tightly written exclusions would strengthen Culpepper’s approach. A shift toward a limited moratorium or performance standards would favor regulated growth instead.

The second signal is action by the Boston Zoning Commission. The commission must consider how the amendment fits the broader zoning code and the city’s development priorities. Its review can expose conflicts between the political concept and practical land-use enforcement.

The commission could recommend revisions before approval. It might seek thresholds for electrical demand, floor area, water use, or backup generation. It could also clarify when a business’s computing equipment remains accessory to another permitted use.

Commission approval with few changes would make the Boston data center ban a meaningful barrier. Extensive exceptions or a broad variance pathway would weaken the prohibition while retaining its headline.

The third signal is Massachusetts’ implementation of Healey’s executive order. State agencies must publish municipal guidance and develop the clean-energy compliance payment structure. The order set a December 31 deadline for key pieces of that framework.

Those rules will determine whether regulated development offers a credible alternative to bans. Developers need to know how additional clean electricity will be measured, how payments will be calculated, and what community agreements must contain.

The state’s 25-megawatt threshold also deserves attention. Boston’s existing Markley facility reportedly draws 30 megawatts, illustrating that a single urban property can exceed it. Future projects could be designed below that threshold while still creating cumulative impacts.

If the state framework produces enforceable protections and clear cost allocation, Boston officials will face pressure to consider conditional approval. If implementation remains vague, a categorical local rule will appear more defensible.

Developers’ behavior provides a related market signal. A company could approach Boston with a smaller facility, a research partnership, dedicated clean generation, limited water use, or a plan to reuse waste heat. Such a proposal would test whether the city truly objects to the land use or mainly to its familiar industrial model.

No application would also be informative. Boston’s property costs and infrastructure constraints might already steer large projects elsewhere. In that case, the amendment would formalize a market outcome while protecting the city from a future change in economics.

The larger AI infrastructure race will continue beyond Boston’s borders. American data center demand is rising, and cloud providers need more capacity. Local governments are increasingly deciding that national technology goals do not override municipal control.

Boston’s choice matters because it is both a technology center and a land-constrained city. A prohibition here would show that enthusiasm for AI software does not guarantee support for the physical infrastructure beneath it.

Readers should watch the text, not only the vote. Does the final rule define data centers precisely? Does it protect legitimate research and network infrastructure? Does it require a transparent public process for exceptions? Those details will determine whether the proposal can survive real applications.

The Boston data center ban remains a proposal, not a completed policy. Its direction is nevertheless clear. Culpepper wants the city to decide the acceptable terms of AI infrastructure before a developer controls the timeline.

That leaves Boston with a direct question: should a dense city reserve an exception for a verifiably low-impact, high-value facility, or is its land and infrastructure already too constrained for that bargain? The hearings, zoning review, and state rules will provide the answer.