When Power Becomes a Credit Condition: Data-Center Underwriting After PJM’s July Filing

The data-center market has spent several years treating power availability as a development constraint. PJM’s late-July 2026 filings make the point more concrete: power can also become an operating and credit condition.

On July 31, PJM filed a proposed Reliability Backstop Procurement and related large-load measures with the Federal Energy Regulatory Commission. The proposal is not final policy, but it puts documented bilateral supply contracts, self-supply arrangements, load registries, and potential load reductions into the same conversation as new large-load growth.

That is a different underwriting environment from one in which a utility service letter is treated as the end of the power analysis. For a data-center project, the practical question is increasingly not only whether a site can connect. It is what the project has contractually secured, when that supply is deliverable, who bears the cost of getting there, and what happens if capacity is scarce.

A current regulatory signal, not a settled rule

The PJM Board’s July 27 decisional letter says new large-load demand is projected to rise by about 70 gigawatts by 2038 while roughly 15 gigawatts of generation has retired in the region since 2022. Those facts describe the regional setting, not a forecast for any one site.

PJM’s proposed Reliability Backstop Procurement would begin in September 2026 and target the capacity shortfall identified in the 2028/2029 auction. Its target would be reduced when documented bilateral contracts or self-supplied new capacity serve new load. Separately, the proposed Interim Resource Adequacy Service would apply to new large loads that, as of June 1, 2027, lack sufficient capacity and cannot otherwise be served at PJM’s reliability standard. Under the proposal, relevant loads could face reductions before pre-emergency load management is deployed.

The important word is proposed. FERC, state authorities, electric distributors, large-load customers, and other stakeholders still matter to the final outcome. A credit memo should not assume the program will take effect exactly as described. But it should recognize the direction of travel: physical power, capacity commitments, and curtailment terms are becoming more visible components of the project-risk package.

The power plan belongs in the sources-and-uses discussion

For a conventional real estate asset, a lender may test leases, sponsor support, construction completion, taxes, insurance, and market vacancy. For a large-load facility, those remain necessary but incomplete. The availability and firmness of power can affect construction timing, customer commencement, equipment commissioning, revenue recognition, operating costs, and the credibility of a completion date.

The underwriting file should therefore distinguish among at least four states that are often blended together in market language:

• A queue position. An interconnection request can be meaningful, but it is not executed, deliverable service. In a January 2026 Dominion submission to PJM, an engineering letter of agreement funds a study but is not firm load, while a capacity letter and a later service agreement carry different reservations, cost, and collateral terms. Names vary by utility; the executed agreement controls.
• A utility commitment. The commitment should be read for conditions precedent, upgrade scope, cost responsibility, milestones, and remedies.
• A capacity arrangement. A bilateral contract or self-supply plan may support a more durable path, but its counterparty, term, credit support, fuel or generation assumptions, and delivery conditions remain finance questions.
• A dispatch and curtailment regime. The project needs to know whether it can be asked to reduce load, what triggers that action, how it is compensated, and how that possibility flows through customer contracts.

This is not a semantic distinction. A project can have a compelling demand story and still face material uncertainty about when it can operate at the required load profile.

Underwrite the date, not just the megawatts

Power is commonly discussed in megawatts. Credit underwriting needs a timeline as well.

Start with the date at which the project must accept service to meet its lease, financing, and equipment obligations. Then test whether every link supports it: interconnection studies, network upgrades, transformer procurement, substation work, generation development, permits, construction labor, and commissioning. A single capacity figure cannot answer that question.

The proposed PJM framework makes the timing issue more visible because it connects new large loads with demonstrated supply arrangements and system reliability. Its June proposal paper also shows how bilateral matching and central procurement were intended to operate in parallel. That is a useful reminder that a project may have multiple paths to capacity, each with a different schedule and allocation of execution risk.

For underwriting, the practical deliverable is a dated power critical path. It should identify the party responsible for each milestone, the evidence supporting the milestone, the outside date, the consequences of delay, and the available fallback. If an assumption cannot be supported by an agreement, permit, utility document, or equipment order, it should be marked as an assumption rather than carried as a fact.

Treat curtailment as an operating-risk scenario

Data-center economics depend on continuity. For some users, an interruption is not a modest revenue variance. It can affect service-level commitments, customer retention, equipment utilization, and the value of the underlying lease.

That makes curtailment terms more than an engineering issue. They belong in base, downside, and severe-stress cases. A project should be able to answer several specific questions:

• Which loads, if any, are interruptible?
• What is the priority of a large load relative to other system actions?
• Can the project run on backup generation, and for how long under applicable permits and fuel contracts?
• Who absorbs lost revenue, replacement power cost, or customer credits?
• Does the customer contract recognize a power-related force majeure or curtailment event?

The goal is not to imply that curtailment is inevitable. It is to ensure that a high-impact event has an assigned contractual owner before it reaches the operating statement.

Grid-upgrade obligations are part of project cost

Recent reporting has also highlighted developers seeking bank-backed facilities to reassure utilities that grid-upgrade obligations will be paid if a project does not proceed. Bloomberg’s August 4 reporting, republished by Yahoo Finance, is a useful market signal, although the individual transactions it describes require project-specific verification.

The broader underwriting implication is durable: any obligation to fund upgrades, post collateral, or reimburse network costs should be treated as part of the capital structure. It needs a source in sources and uses, a draw schedule, a collateral treatment, and a downside case. It should not sit outside the model as an informal development contingency.

A practical credit-file checklist

For investment committees and lenders, a short power appendix can be more useful than a generic market paragraph. At a minimum, it should include:

1. The requested load, expected load profile, location, and required in-service date.
2. The executed documents and the conditions that remain outstanding.
3. A milestone schedule for interconnection, network upgrades, equipment, and commissioning.
4. The expected capital cost, collateral, deposits, and party responsible for overruns.
5. Any bilateral supply, self-supply, or backup-generation arrangement, including counterparty credit and term.
6. Curtailment rights, compensation, customer remedies, and operating-continuity plan.
7. A downside case that delays service, increases upgrade cost, or reduces available load.

The relevant discipline is the same one applied to any infrastructure-dependent real estate asset: separate signed rights from planned rights, and separate timing assumptions from completed work.

The underwriting conclusion

PJM’s pending proposals may change before they are implemented, and their direct applicability is regional. Their importance reaches further. They make visible a condition that data-center underwriting has to confront in many markets: reliable power is not simply a local utility attribute. It is a bundle of physical infrastructure, contractual commitments, regulatory processes, costs, and operating contingencies.

The better credit file will not reduce that complexity to a single power-available flag. It will show how the project obtains capacity, what evidence supports the delivery date, what risks remain with the developer or customer, and how the asset performs if the plan slips. That is how power becomes a credit condition rather than a late-stage surprise.