U.S. Data Center Restrictions Are Rising. The AI Infrastructure Boom Isn’t Ending—It’s Changing Geography.

Are States Restricting Data Centers? What the Backlash Really Means for America’s AI Infrastructure and the Manufacturers Building It

Artificial intelligence is driving one of the largest infrastructure buildouts the United States has seen in decades. But as hyperscale and AI data centers demand unprecedented amounts of electricity, water, land and supporting infrastructure, states and local governments are beginning to push back.

Arizona is the latest example.

On August 31, 2026, Arizona Attorney General Kris Mayes called for a statewide pause on the approval and construction of new data centers, particularly hyperscale AI projects. She urged Arizona lawmakers and Governor Katie Hobbs to establish a temporary pause while the state develops a more comprehensive plan for managing data-center growth.

The announcement generated understandable headlines about another state potentially moving against data centers.

But there is an important distinction.

Arizona has not enacted a statewide construction ban.

And across the United States, most of the restrictions being discussed are not outright bans either.

Instead, the data-center industry appears to be entering a new phase in which developers must demonstrate that sufficient power, transmission, water and infrastructure are available—and increasingly pay more of the cost required to provide them.

For the manufacturers supplying the equipment behind these facilities, that distinction is critical.

The data-center boom may not be ending.

It may be changing geography, infrastructure requirements and the way the industry builds.

The Data Center Backlash Is Real

The regulatory movement surrounding data centers has expanded quickly.

One nationwide tracker identified state-level data-center policy action in 39 states as of August 2026, including tax-incentive rollbacks, siting restrictions, regulatory changes, proposed moratoriums and executive actions. It identified 19 states with restrictions in force, three states with active pauses or executive moratoriums, and nearly 300 local actions involving cities, counties, townships and other jurisdictions.

The National Conference of State Legislatures reported in July 2026 that lawmakers in approximately 15 states were considering some form of data-center development moratorium.

But these actions vary enormously.

There is a major difference between:

• an outright construction moratorium;

• a temporary permitting pause;

• a local zoning restriction;

• a utility interconnection delay;

• a requirement that a data center pay for grid upgrades;

• elimination of a tax incentive;

• additional water requirements; and

• a community rejecting one particular project.

Calling all of these measures “data-center bans” creates a misleading picture of what is actually happening.

In most cases, governments are not saying that data centers can no longer be built.

They are asking a different question:

Who should pay for the infrastructure required to support them?

Arizona: What Actually Changed?

Arizona illustrates the distinction particularly well.

The August 31 announcement from Attorney General Mayes is a proposal for a statewide construction pause—not a pause that is currently in force.

Arizona has, however, already changed its economic treatment of new data centers.

Beginning July 1, 2026 and continuing through June 30, 2029, the Arizona Commerce Authority cannot accept new applications for the state’s computer-data-center tax-relief program. Previously approved projects retain their existing treatment.

That is a meaningful policy change, but it does not prohibit construction.

Arizona regulators are also moving toward a principle that could become increasingly common across the country:

Growth pays for growth.

The basic idea is that very large electricity users should pay the infrastructure costs created by their projects rather than shifting those costs to existing households and businesses.

That changes the economics of data-center development, but it does not necessarily eliminate demand.

Instead, it puts greater pressure on developers to secure power, fund infrastructure and select locations where projects can actually be delivered.

New York Has Gone Further

New York provides a much clearer example of an actual statewide pause.

On July 14, 2026, Governor Kathy Hochul issued an executive order establishing a one-year moratorium on new hyperscale data centers while the state develops a regulatory framework addressing ratepayer, environmental and community impacts.

New York’s action represents one of the strongest statewide responses to the industry’s rapid expansion.

Other states are taking different approaches.

Rather than prohibiting construction, states are increasingly reconsidering tax incentives, establishing special electricity rates for very large loads, imposing additional permitting requirements, protecting water resources and requiring developers to finance electrical infrastructure associated with their projects.

That distinction matters because the underlying demand for computing capacity remains extraordinarily strong.

Demand Is Still Outrunning Supply

If regulatory resistance were causing the overall data-center market to collapse, the effect should begin appearing in vacancy, leasing and construction numbers.

The latest market data shows almost the opposite.

According to CBRE’s H1 2026 North America Data Center Trends report:

• Primary-market supply increased 33.7% year over year.

• Capacity under construction increased 24.8% to a record 7,481 MW.

• More than 80% of that construction was already preleased.

• Vacancy fell to a record-low 1.4%.

• Less than 1,500 MW of future primary-market capacity remained available for preleasing—roughly six months of supply at the current absorption rate.

JLL’s broader North American market analysis tells a similar story.

It reported approximately 66 GW of data-center capacity under construction, with 95% already committed, and said vacancy remained around 1%. JLL also reported a record 25 GW of absorption during the first half of 2026.

Different research firms use different methodologies and market definitions, so those absolute capacity figures should not be directly compared.

But they point to the same conclusion:

The problem facing the data-center industry is not a lack of demand. It is the difficulty of supplying that demand.

Electricity Is Becoming the Real Constraint

For decades, fiber availability, land and real estate were among the primary considerations determining where data centers were developed.

Increasingly, the first question is much simpler:

How many megawatts can you deliver—and when?

A June 2026 update from Lawrence Berkeley National Laboratory estimates that data centers could account for approximately 11.8% of total U.S. electricity consumption by 2030.

Its modeled scenarios range from approximately 9.5% to 15.3% of total U.S. electricity consumption.

Loads of this magnitude have implications far beyond the individual data-center building.

A major AI campus may require:

• additional electrical generation;

• new transmission capacity;

• substations;

• transformers;

• medium- and low-voltage switchgear;

• backup generation;

• UPS systems;

• energy storage;

• cooling infrastructure;

• control systems; and

• extensive supporting mechanical and electrical equipment.

In effect, the modern hyperscale AI data center is increasingly becoming both a computing project and a power-infrastructure project.

Federal regulators have recognized the challenge.

In June 2026, the Federal Energy Regulatory Commission directed all six regional grid operators under its jurisdiction to justify or reform the rules governing how data centers and other very large electricity users connect to the grid. FERC specifically cited the need to accelerate large-load connections while protecting grid reliability and existing ratepayers.

That development may ultimately prove more important to the future of the industry than individual local moratoriums.

The Geography of the Data Center Boom Is Changing

Power constraints are also changing where data centers are being built.

Northern Virginia remains the largest established data-center market in North America. But even there, available land, permitting and power availability have become increasingly difficult.

CBRE reported that Atlanta overtook Northern Virginia in H1 2026 for total capacity under construction.

Atlanta reached approximately 2,882 MW under construction, while Northern Virginia had approximately 2,420 MW. Northern Virginia nevertheless remained extremely tight, with vacancy of only 0.2%.

JLL describes an even broader geographic shift.

According to its midyear 2026 report, 77% of North American data-center capacity under development is now located in what it calls frontier markets.

West Texas has been one of the largest beneficiaries, while Ohio, Louisiana, Indiana and the Carolinas are also attracting substantial development.

Many of these locations had relatively little data-center capacity a decade ago.

The industry is effectively beginning to follow available power.

This suggests that restrictions in one established market do not necessarily destroy demand.

They can redirect it toward another jurisdiction.

Demand Destruction or Demand Redirection?

That is perhaps the most important distinction in the current data-center debate.

If an AI infrastructure project planned for one market encounters a five-year power delay, restrictive zoning or an unfavorable cost structure, the underlying computing requirement does not necessarily disappear.

The developer may:

• move the project;

• divide the project among several locations;

• choose a secondary market;

• secure dedicated power generation;

• construct its own electrical infrastructure;

• use modular infrastructure;

• redesign the cooling system; or

• enter into a different utility arrangement.

That is demand redirection, not necessarily demand destruction.

The manufacturing requirements follow the project.

Transformers are still required.

Switchgear is still required.

Generator systems are still required.

Cooling systems are still required.

Power-distribution equipment is still required.

Enclosures, frames, cabinets, panels, skids and fabricated assemblies are still required.

They may simply be required somewhere else.

The Data Center Is Becoming a Power Project

One of the biggest changes occurring in AI infrastructure is the growing relationship between data centers and dedicated power generation.

Developers increasingly are examining or deploying combinations of:

• natural-gas generation;

• nuclear power;

• renewable generation;

• battery energy storage;

• fuel cells;

• microgrids; and

• behind-the-meter generation.

This changes the physical supply chain surrounding data centers.

A large AI campus may increasingly resemble:

Generation → Transmission → Substation → Transformer → Switchgear → UPS → Distribution → Computing → Cooling

Every additional layer creates requirements for physical infrastructure.

That means the manufacturing opportunity surrounding AI extends much farther than racks of servers.

A Manufacturing Bottleneck Is Developing Behind the AI Boom

Building the data center is only part of the challenge.

Manufacturers must build the equipment going inside it.

And several critical equipment categories continue to carry substantial lead times.

SourceBlue’s Q2 2026 Cost Index reported estimated lead-time ranges of:

• 26-58 weeks for low-voltage switchgear

• 28-48 weeks for medium-voltage switchgear

• 28-66 weeks for generators

• 16-42 weeks for UPS equipment

• 12-60 weeks for water-cooled chillers

• 8-78 weeks for air-handling units

Conditions have improved in parts of the supply chain, but those ranges show why manufacturers are continuing to expand production capacity.

The bottleneck is no longer simply acquiring GPUs or securing real estate.

It can be the electrical cabinet.

The welded frame.

The generator enclosure.

The cooling-equipment assembly.

The transformer housing.

The power-distribution enclosure.

The equipment skid.

Or the manufacturing capacity needed to produce all of them fast enough.

Equipment Manufacturers Are Responding

Major electrical, power and cooling manufacturers have announced significant U.S. capacity investments as they respond to data-center and AI infrastructure demand.

Companies including Eaton, Siemens, Schneider Electric, Vertiv, ABB, Hitachi Energy and others have expanded or announced additional North American manufacturing capacity for electrical distribution, transformers, switchgear, power equipment and supporting infrastructure.

Cooling manufacturers are expanding as well.

The reason is straightforward.

When data-center demand grows faster than equipment-production capacity, the industry’s constraint moves upstream into the factories supplying the project.

This creates another question for equipment OEMs:

How much of the manufacturing process really needs to remain inside their own factory?

Why Contract Manufacturing Matters

A manufacturer of switchgear, UPS equipment, cooling systems, transformers, generator packages or other data-center equipment typically has processes that represent its core technical value.

Those may include:

• electrical engineering;

• proprietary controls;

• thermal technology;

• electrical integration;

• final configuration;

• testing;

• validation; and

• certification.

But the same product may also contain substantial quantities of fabricated metal.

Examples can include:

• cabinets;

• doors;

• sheet-metal panels;

• electrical enclosures;

• welded frames;

• structural supports;

• generator enclosures;

• battery cabinets;

• cooling-equipment housings;

• mechanical assemblies;

• equipment bases;

• skids; and

• coated or painted fabricated components.

When internal fabrication becomes a production bottleneck, an OEM does not necessarily need another complete factory.

It may need additional qualified manufacturing capacity.

By moving appropriate fabrication, finishing and assembly operations to an outside manufacturing partner, OEMs can preserve valuable internal factory capacity for the engineering, electrical integration and testing processes that are harder to outsource.

Fewer Manufacturing Handoffs Can Also Mean Faster Throughput

Increasing capacity is only part of the issue.

The number of manufacturing handoffs can also affect lead time.

A component might otherwise move from one supplier for fabrication, to another for welding, to another for coating, and perhaps to another location for assembly or packaging.

Each handoff can create:

• transportation;

• scheduling delays;

• additional work-in-process inventory;

• packaging;

• receiving and inspection;

• supplier coordination;

• handling; and

• administrative complexity.

An integrated contract-manufacturing model can combine multiple stages of that value stream.

Prince Manufacturing, for example, provides contract-manufacturing capabilities that can include metal fabrication and forming, welding, metal finishing, assembly, packaging, kitting, sequencing and shipping. Depending on project requirements and the appropriate Prince location or qualified partner, processes can be consolidated into a more integrated manufacturing program.

The concept is straightforward:

Form → Finish → Assemble → Package → Sequence → Ship

Instead of treating each process as an isolated purchase order, the objective is to manage more of the manufacturing value stream under one roof.

There is documented evidence that this type of consolidation can materially affect cycle time.

What This Means for Data-Center Equipment OEMs

The next phase of the data-center boom may place a premium on manufacturing flexibility.

Demand is growing.

Projects are moving geographically.

Power infrastructure is expanding.

Cooling requirements are becoming more sophisticated.

Electrical equipment manufacturers are adding capacity.

At the same time, developers are demanding shorter schedules because access to power itself has become valuable.

That combination can create enormous pressure upstream.

OEMs supplying switchgear, power distribution, UPS systems, backup generation, energy storage, cooling systems and modular infrastructure may increasingly need ways to add production capacity without waiting for a new factory to be designed, permitted, equipped and staffed.

Contract manufacturing can provide one path.

Instead of outsourcing a finished proprietary product, an OEM can selectively outsource the portions of the manufacturing value stream that consume capacity but do not necessarily represent its core intellectual property.

That may include fabrication, welding, finishing, coating, assembly, packaging or sequenced delivery.

The result can be additional manufacturing throughput without transferring the OEM’s core engineering or technology.

The Data-Center Boom Isn’t Over. The Rules Are Changing.

The regulatory backlash surrounding data centers is real.

States and municipalities are questioning whether local communities should absorb the cost of electricity infrastructure, water consumption, land use and other impacts generated by very large projects.

More restrictions are likely.

Some projects will be delayed.

Some projects will be rejected.

Some locations will become substantially more difficult or expensive to develop.

But current market data does not indicate that the fundamental demand for digital and AI infrastructure has disappeared.

North American data-center vacancy remains near historic lows. Construction remains at record levels. Most new capacity is committed before completion. Equipment manufacturers continue expanding production.

The more likely transformation is that development increasingly follows locations capable of supplying power, infrastructure and regulatory certainty.

That shift has significant consequences for manufacturers.

As billions of dollars of data-center investment move into new markets, the electrical, power-generation, cooling and mechanical equipment supporting those facilities must move with it.

And behind virtually every megawatt of new computing capacity is a substantial amount of manufactured physical infrastructure.

Need Additional Manufacturing Capacity for Data-Center Equipment?

For OEMs manufacturing switchgear, electrical distribution equipment, UPS systems, generator packages, cooling equipment, energy-storage systems, enclosures, cabinets, skids and other data-center infrastructure, production capacity can become just as important as demand.

Prince Manufacturing can help relieve manufacturing bottlenecks by taking appropriate fabricated-metal content, finishing and assembly work outside of constrained OEM facilities.

Prince’s contract-manufacturing capabilities include metal fabrication, sheet-metal forming, stamping, MIG/TIG/spot welding, powder coating, E-coating, liquid painting, assembly, packaging, kitting, sequencing and shipping, with manufacturing operations in the United States and Mexico. Specific processes and plant capabilities should be matched to each program’s technical requirements.

Instead of managing separate suppliers for fabrication, finishing and assembly, qualifying programs can consolidate multiple processes through one manufacturing partner:

FORM → FINISH → ASSEMBLE → PACKAGE → SEQUENCE → SHIP

The objective is simple: add manufacturing capacity, reduce supplier handoffs, shorten the value stream and help data-center equipment OEMs get more product out the door faster.

If manufacturing capacity or fabricated-metal lead time is becoming a constraint on your data-center infrastructure program, Prince Manufacturing can review your drawings, components or RFQ package and determine where an integrated contract-manufacturing solution may help. Contact Us.