US Data Center Development: Power, Permits and Deliverability

The next phase of US data center development demands a more precise approach to site selection. AI creates demand for computing infrastructure, but each project still depends on a local chain of decisions: a utility connection, a permitted design, a viable electricity contract and a community willing to host it.

For developers and investors, this means assessing opportunities at the level of the parcel, the utility service territory and the delivery phase. A national growth forecast cannot establish when a particular building will receive power.

PowerlandMap’s US supply snapshot illustrates the scale of the work ahead. As of 5 November 2026, our published, publicly verified inventory, excluding demo records, contains 1,566 US records. Of these, 180 are under construction and 315 are in development: a combined pipeline of 495 records, or 31.6% of the total.

What PowerlandMap’s US data shows

These figures show why pipeline analysis needs to distinguish maturity. An operating asset, a construction project and an early development opportunity have different dependencies. Adding them together without stage definitions can make a market appear more deliverable than the evidence supports.

The 495 pipeline records are a research universe. They do not establish how many projects will open on time, how much capacity is uncommitted, or which customers can actually be served. Those questions require project-level evidence.

Power availability must be demonstrated phase by phase

A substation near a site is useful context. It does not prove that the utility can connect the required load, at the required date, with the required redundancy.

A credible US data center development plan should specify the first energisable tranche, the works needed to deliver it and the conditions attached to subsequent phases. It should distinguish requested load from reserved or contracted supply, and electrical capacity from IT capacity.

The relationship between the utility’s schedule and the customer’s ramp-up is especially important. If an initial building can be energised but later phases depend on transmission reinforcement, those phases need separate schedules and investment assumptions.

AEP Ohio provides a concrete example of why demand requests and contractual commitments must be tracked separately. In its February 2026 update, the utility said approximately 30,000 MW of requests affected by its new tariff had progressed to 13,022.7 MW of paid engineering studies, with 5,642 MW covered by signed binding contracts as of 12 February.

Those figures describe a particular group of requests. AEP Ohio also reported 12,219 MW contracted before the tariff, taking total contracted demand to approximately 17,861 MW, phased through 2035. The distinction matters: the 5,642 MW figure is neither the utility’s entire data center commitment nor evidence that every remaining request has disappeared.

Source: AEP Ohio’s data center demand update.

Electricity contracts are part of the development strategy

Grid access has a commercial structure as well as an engineering schedule. AEP Ohio’s data center tariff illustrates this through minimum demand charges, financial security requirements and long-term commitments. Its framework includes a ramp period of up to four years followed by eight years of obligations. Specific terms depend on the customer and agreement.

For a developer, these provisions affect the cost of committing to a site before tenant demand has fully materialised. They should be assessed alongside land payments, connection works and building expenditure.

A project model should therefore test what happens if energisation, construction or customer occupancy moves by twelve months. The central question is which costs remain payable, who carries them and whether the project can still finance its next phase.

Source: AEP Ohio data center tariff information.

Local acceptance belongs in the site selection model

The political dimension of AI infrastructure becomes tangible through local land-use decisions. Noise, equipment placement, water requirements, visual impact and electricity affordability can all influence project design and approval.

Fairfax County offers a specific example. Its data center zoning changes, adopted in September 2024, introduced residential separation requirements and noise studies before and after construction. The rules include a 200-foot building setback in specified circumstances and additional requirements for equipment placement or screening, subject to the ordinance’s conditions and exceptions.

These requirements influence developable area, equipment layout and mitigation costs. A large parcel can become a materially different opportunity once the permitted design is tested.

Source: Fairfax County’s data center zoning amendment.

Community engagement should begin while the design can still change. Developers need a credible account of operating impacts, local benefits and responsibilities. Generic promises about AI growth will not answer a neighbour’s question about generator testing or a council’s concern about infrastructure costs.

Competition must be assessed within the same delivery window

Two projects in the same metropolitan area may serve different requirements. One may offer an earlier first phase, another a larger expansion path, and a third stronger fibre connectivity or a different cooling configuration.

The relevant comparison is therefore specific: projects that can serve a compatible workload, within a comparable delivery window, under credible power and permitting assumptions.

For AI infrastructure, screening should also address rack density, liquid-cooling readiness, network requirements and the implications of the intended workload. A training campus and a latency-sensitive deployment may prioritise locations differently.

PowerlandMap’s approach combines project-stage information with mapped electricity infrastructure and source evidence to help build that shortlist. Its methodology keeps announced, requested, reserved, contracted and deliverable power concepts distinct. Mapped infrastructure supports investigation; utility confirmation remains necessary.

A more precise development brief

Before committing substantial development capital, the project team should be able to document:

• The first power tranche, its delivery date and the evidence supporting both.
• The permissions and physical constraints governing the proposed design.
• The financial obligations attached to the utility connection and load ramp.
• The competing projects targeting the same customers and delivery window.
• The technical fit between the proposed facility and the customer requirement.
• The conditions for expansion, including what remains unconfirmed.

This is where market intelligence becomes useful to development execution: identifying missing evidence early, comparing opportunities on a consistent basis and directing diligence to the assumptions that could change the decision.

The US opportunity remains substantial in the inventory we track. Converting it into operating infrastructure requires increasingly detailed work. The strongest development proposition is a defined phase, supported by evidence, that can serve a real customer on an achievable timetable.

Explore PowerlandMap’s live market coverage or request a scoped evaluation to assess your target market and site shortlist.

Data note: PowerlandMap figures are a dated inventory snapshot, not an estimate of the entire US market. Stage counts do not certify available capacity or future completion. Cover and substation visuals are AI-generated illustrations, not photographs of named facilities.