How Buyers Should Test a Data Center Power Claim

A credible data center power claim is one a buyer can trace through utility status, physical delivery design, commercial responsibility, and a dated path to service. Treat stated megawatts as a starting point, not a selection criterion; a queue position, conditional study, and energized capacity are very different things.

That sounds obvious, but it gets lost when an RFP asks every provider for “available power” and everyone answers with a large number. The buyer who asks better follow-up questions gets a much clearer view of deployment risk before legal, network, and design teams spend months on a deal that cannot close on the required date.

What does “available power” actually mean?

Ask the operator to put its answer into one of four buckets:

• Energized and uncommitted: Capacity is already delivered to the site and can be allocated under the proposed contract.
• Energized but needing internal buildout: The utility supply is live, while substation, distribution, hall, or customer-specific equipment still needs to be installed.
• Contracted or approved, with work remaining: The utility has committed to a delivery path, but transmission, substation, or feeder work must be completed.
• Requested or conditional: The project has an application, study result, queue place, land position, or preliminary utility discussion, but not a fully executable power-delivery path.

None of the latter three buckets automatically disqualifies a site. Many large requirements cannot be met with already-energized capacity. The issue is whether the operator represents the category accurately and whether your deployment schedule can tolerate the remaining dependencies.

A provider saying “we have 100 MW available” may mean it has 100 MW behind an operating meter. It may also mean its campus has a long-term ambition supported by a utility application. Those statements should never be treated as equivalent in a shortlist.

Which documents should buyers ask to see?

You do not need to demand a utility’s confidential file. You do need enough evidence to understand where the project sits and who owns the next action.

For a near-term deployment, request a concise power evidence pack: the serving utility and point of delivery; the current energized capacity; the capacity already committed to other customers; the interconnection or service-agreement status; the critical construction milestones; and the assumptions that could move the in-service date. Ask for a single-line diagram at an appropriate level of detail, too. It quickly shows whether a claimed utility feed, on-site generation scheme, and redundant topology are real design elements or sales shorthand.

Then ask a question that often exposes weak answers: What must happen outside your fence line before my contracted load can be served?

A good operator can name the substation, feeder, transmission upgrade, utility approval, or fuel arrangement involved. They can also explain who funds it, who controls it, and what happens if it slips. A vague answer about “working closely with the utility” is not evidence.

This distinction is getting sharper in Texas. ERCOT has begun sending Batch Zero verification requests for information to most conditionally included large-load interconnection applicants as part of its directed audit process (ERCOT). If an operator cites ERCOT progress in a sales cycle, buyers should ask whether the project is conditional, what verification has been requested, and what that means for the stated delivery date. A queue milestone is useful. It is not the same as a delivered-power commitment.

How should you test the promised delivery date?

Build the date backward from the moment your equipment must accept load. Do not start with the provider’s headline opening date.

For each proposed phase, get the date and owner for utility energization, customer-space readiness, electrical commissioning, network turn-up, and the point at which your cabinets or suites can be handed over. This separates a building completion date from a usable deployment date. In practice, the latter is what your infrastructure team, cloud group, or enterprise tenant cares about.

Also identify the longest pole in the tent. It could be an external transmission upgrade. It could be procurement of major electrical equipment. It could be the customer’s own liquid-cooling design decision. If the seller cannot identify the critical path, they are not managing the schedule at the level a serious buyer needs.

Large-load reliability rules are also becoming a practical diligence item rather than a utility-side technicality. PJM proposed requirements after a July event in which nearly 4,000 MW of Northern Virginia data-center load disconnected, creating voltage and frequency challenges during the transition to backup generation (PJM). For buyers in PJM territory, that makes it sensible to ask how the facility sequences backup generation, coordinates with the grid operator, and handles a sudden loss of utility supply. “We have generators” is not a complete answer.

Who pays when grid work gets more expensive?

This is where commercial and technical diligence meet. A quoted colocation rate may look stable while the underlying power arrangement contains pass-throughs, capacity charges, curtailment provisions, or customer obligations for future upgrades.

Ask the provider to explain, in plain language:

• What portion of your price is fixed, indexed, or passed through.
• Whether utility upgrade costs are included, capped, or subject to change.
• What events permit a change in the service date or available contracted load.
• Whether the provider can curtail load, and under what conditions.
• What remedy applies if it cannot deliver the contracted capacity on time.

The last point deserves more attention than it usually receives. A service-level credit designed for a brief outage does not compensate for a delayed deployment that forces you to extend a temporary site, defer revenue, or scramble for capacity elsewhere. Your deal team should match remedies to the actual cost of delay.

Regional price exposure should be part of this conversation as well. Pexapark, citing Monitoring Analytics, reported that data-center growth accounted for $10.48/MWh, or 9%, of PJM wholesale power costs in the first seven months of 2026 (Pexapark). Buyers do not need to become power traders. But they should understand which parts of a proposed energy bill can move and how much notice they will get.

When is flexible power a good option?

Flexible or non-firm power can be a smart commercial choice if your workload can genuinely shift, pause, or run from backup resources at defined times. It is a poor choice when the application requires continuous, firm delivery and the operating team has no practical way to shed load.

Do not reject flexibility because it sounds inferior. Instead, quantify it operationally. Ask how often curtailment may occur, how much notice is provided, what load must come off, and whether backup generation can cover the gap without compromising your resilience posture. Ask whether the flexibility is a temporary bridge to firm service or a permanent condition of the contract.

MISO is currently considering an interim flexible-large-load service and a contract-demand construct for large loads, with stakeholder feedback requested this month (MISO). That is a useful reminder that flexible service is becoming more formalized in some markets. For the right buyer, it may widen site options. For the wrong workload, it simply moves risk from the provider’s development plan onto your operations team.

How should brokers and enterprise teams score competing sites?

Use a power-confidence score alongside the usual price, location, network, and facility criteria. The score should reward evidence, not ambition.

One practical approach is to mark each candidate on current deliverable load, external dependencies, schedule certainty, commercial exposure, backup and curtailment design, and the operator’s willingness to provide documentation. Keep the assessment qualitative if that suits your procurement process. What matters is that a site with conditional capacity does not quietly tie with a site that can deliver now just because both fit into the same “MW available” field.

Make the operator explain the weak spots in its own plan. The strongest sellers will do it without being defensive, because they know an honest risk register creates a better deal and fewer surprises after signature.

Common questions

Is a utility queue position enough to select a data center site?

No. A queue position can be an important early milestone, but it does not by itself prove a delivery date, funded upgrades, or firm service. Ask what approvals and construction work remain before your load can be energized.

Should we only consider already-energized capacity?

Not necessarily. Requiring energized capacity may eliminate viable sites for a large or future deployment. The better approach is to align the remaining power dependencies with your schedule and contract protections.

What is the biggest red flag in a power presentation?

A large capacity number with no clear definition is the usual problem. If the provider cannot distinguish current energized load from future planned capacity, pause the shortlist until it can.

Can non-firm power work for an AI deployment?

It can, if workloads can be managed around curtailment and the economics justify it. It is much less suitable when compute must run continuously and an interruption would disrupt a customer commitment or a tightly scheduled training run.

Where this leaves you

Power diligence should be a structured part of site selection, not a late-stage engineering check. Put every claim into a delivery category, trace the dependencies, and make the contract reflect the cost of a miss. GridReach helps data center and energy companies turn expertise like this into qualified pipeline.