How Buyers Vet On-Site Power Plans

A credible on-site power plan can improve a data center’s path to service, but it does not remove buyer due diligence. Buyers should verify the fuel path, permit status, construction sequence, operating responsibility, and what happens when the plant cannot carry the promised load.

The mistake is treating behind-the-meter generation as a simple alternative to utility power. It is a separate operating system attached to your facility decision, with its own dependencies, contracts, emissions exposure, maintenance plan, and failure modes.

Why are buyers asking harder questions about on-site power?

Because the concept has moved from a speculative slide in a pitch deck to real project infrastructure. In West Texas, Kodiak Gas Services has agreed to supply 76 MW of behind-the-meter baseload power to a data center, using natural-gas reciprocating units that it expects to begin deploying in Q4 2026 and scale into Q1 2027 (Kodiak Gas Services). That is a useful example of what buyers now need to inspect: not just a stated capacity number, but equipment type, deployment timing, operator, and fuel source.

For an enterprise infrastructure team, a broker, or a site-selection adviser, the practical question is not, “Does this campus have generation?” It is: “What workload can this facility support, on what date, under what operating conditions, and whose obligation is it if that answer changes?”

That distinction matters especially when the buyer has already committed to GPU deliveries, migration dates, or a lease expiry at another site. A vague power story might be acceptable during initial market research. It is not acceptable when a deal is moving from technical validation to legal review.

What should a power claim include before a site makes the shortlist?

Ask the operator for a short, dated power evidence package. It does not need to disclose every commercial term, but it should make the claim testable.

At minimum, the buyer should be able to see:

• The proposed operating capacity for the buyer’s phase, rather than only the campus’s eventual headline capacity.
• The source of electricity and, for on-site generation, the fuel-delivery arrangement.
• The status of critical permits, equipment orders, interconnection work, and site construction.
• The expected commercial-service date and the assumptions behind it.
• The division of responsibilities among the facility owner, generator operator, utility, fuel supplier, and tenant.
• The backup, maintenance, and failure-response design.

This is not needless bureaucracy. Each item answers a different question. A gas supply arrangement does not prove that turbines are permitted. A permit does not prove that equipment is on order. An equipment order does not prove that the operator has a workable maintenance model.

Buyers should also request that the operator separate firm commitments from preliminary development work. A developer may have a persuasive long-range plan, but the buyer’s deployment decision is usually about a specific hall, a specific reservation date, and a specific amount of usable capacity.

How do you test the fuel and generation chain?

Follow the physical chain, not the marketing diagram. If the facility depends on gas-fired generation, ask where the gas enters, who has contracted for transport and supply, and what infrastructure must be completed before the generators can run at the required load.

Woodway Energy Infrastructure recently announced definitive agreements for a 22-mile, 30-inch gas pipeline serving a behind-the-meter generation project for hyperscale data-center development (Woodway Energy Infrastructure). The point for buyers is not that every site needs a dedicated pipeline. It is that fuel infrastructure can be a real project workstream, not a footnote beneath the phrase “on-site power.”

Ask for the constraints in plain language. Is fuel supply firm or interruptible? Is there a single delivery path? What happens during a curtailment, pipeline maintenance event, or extreme weather event? Can the generation fleet serve the entire contracted IT load, or only a portion? And how much fuel is stored or otherwise available if normal delivery is disrupted?

You do not need to become a gas-market specialist. You do need an answer from someone accountable for operating the system, not only someone selling capacity.

Are permits merely a local issue?

No. Permits are a schedule, community, and reputational issue all at once. They may affect what can be built, when it can operate, what environmental controls are required, and whether a project’s assumptions remain intact after public scrutiny.

South Carolina regulators recently approved an air permit for roughly 450 MW of gas generation supporting a data center under construction in Spartanburg (Spartanburg Herald-Journal via AOL). That is a reminder that generation development has a permitting track distinct from the data-center shell. Buyers should ask which approvals have been secured, which remain pending, and whether the service-date commitment depends on an unresolved approval.

Regulatory expectations can also change after a site is identified. California enacted data-center-related measures addressing matters including electricity cost allocation, energy procurement, water-use disclosure, and environmental review; the governor’s office said the package requires data centers to pay grid-upgrade costs and comply with applicable energy procurement requirements (Office of the Governor of California). Even if your project is elsewhere, it is sensible to ask how the operator monitors policy risk and allocates the cost of compliance.

Honestly, “we’ll handle permitting” is not enough. A strong operator can explain the remaining path, its critical dependencies, and what contingency exists if a milestone slips.

Who operates the plant after the sales team leaves?

This is where many otherwise polished proposals get thin. On-site generation should have a named operator, an operations model, defined maintenance windows, spare-parts planning, monitoring, and an escalation process that lines up with the facility’s service-level commitments.

Buyers should press on the awkward scenarios:

• A unit trips during a high-density commissioning window.
• Maintenance is required when the tenant’s load cannot be reduced.
• A fuel or equipment issue limits output below the contracted level.
• Utility supply and on-site generation need to transfer or run in parallel.

The answers should identify who makes decisions, who communicates with the tenant, and what remedy applies. “Our engineering team has it covered” does not tell procurement or legal enough to assess the risk.

For AI deployments, load behavior belongs in this conversation too. The AI Energy Management Alliance launched by Google, NVIDIA, and Emerald AI is promoting workload shifting, storage dispatch, paired generation, and responses to grid contingencies as ways to make data-center loads more grid-responsive (NVIDIA). That may create useful options, but buyers should not assume their workloads are flexible merely because the facility says flexible operation is possible. Confirm which workloads can move, who authorizes a change, and whether performance commitments survive it.

When should buyers bring these questions into the deal?

Earlier than most do. During the first shortlist, use a concise questionnaire to eliminate sites whose claims cannot be substantiated. During technical due diligence, review the actual documents and operating assumptions. Before signature, make sure the commercial agreement reflects the promises that mattered in selection.

That sequence avoids a common late-stage problem: the sales presentation describes available power, but the contract only commits to a future best effort. If an on-site plant is central to your decision, its delivery milestones, capacity definition, availability expectations, and remedies deserve explicit treatment.

A buyer does not need perfect certainty before selecting a site. No major infrastructure project offers that. But you should know exactly which dependencies you are accepting, which party owns them, and what recourse you have if they do not resolve on time.

Common questions

Does behind-the-meter power mean the utility is no longer relevant?

Usually, no. The utility relationship may still matter for interconnection, auxiliary loads, backup arrangements, exports, or the future transition to grid supply. Ask the operator to map where utility dependency remains rather than presenting the site as wholly independent.

Should we reject a site if every permit is not final?

Not automatically. What matters is whether outstanding approvals are material to your required service date and whether the operator can show a credible path through them. Treat pending approvals as an identified schedule risk, then decide whether the contract and contingency plan address it adequately.

Can a PPA prove that our power will be available?

A PPA can be important evidence, but it is only one part of the chain. You still need to understand delivery arrangements, generation or grid constraints, facility readiness, and the terms governing your specific contracted capacity.

What should our broker ask that we might miss?

Ask the broker to compare sites using the same definitions of capacity, service date, density, and redundancy. A broker can also force a clean distinction between signed commitments, permits in hand, equipment being installed, and concepts still under development.

What to do with this

Add a power-proof review to your shortlist process before your team gets emotionally invested in a location or commercial proposal. The best operators will welcome a disciplined review because it separates a real operating plan from a broad capacity claim.

GridReach helps data center and energy companies turn expertise like this into qualified pipeline.