Why Buyers Ask About Curtailment Now

Flexible demand is becoming a real shortlist issue because it can affect a facility’s power economics, operating model, and expansion rights. Buyers should not reject curtailment outright, but they should force a precise discussion of when it applies, who controls it, and what workload can safely move.

For years, a customer asking about utility programs could sound like a side conversation after the power and price discussion. That has changed. Large new-load agreements are increasingly being paired with customer-funded infrastructure and explicit flexibility commitments. In Georgia, for example, the approved agreement for OpenAI’s Effingham County project includes project-specific infrastructure costs and up to 1,000 MW of flexible demand response (Georgia Power).

That does not mean every colo buyer will be asked to curtail at that scale. It does mean sophisticated buyers will assume power is a commercial product with conditions, rather than a binary facility attribute.

Why has curtailment reached the shortlist?

The simple answer is that available utility capacity is scarcer and more conditional than the headline megawatt number suggests. CBRE reported that more than 80% of the construction pipeline in its primary North American markets was preleased, while power and infrastructure constraints were extending completion schedules (CBRE).

When capacity is constrained, utilities and large customers look for ways to make new demand less punishing during stressed periods. Flexible-load provisions can be one route to an earlier service agreement or a more workable grid arrangement. They can also be a way for the utility to protect the broader system without treating a new data center as completely inflexible around the clock.

Buyers understand that trade. What they dislike is ambiguity.

A site selector who hears “we have flexible power” will now ask whether that means an attractive commercial option, an unpriced operating obligation, or a power commitment that is less firm than the sales deck implies. If your team cannot answer that cleanly, the site may survive the first call but lose credibility during technical diligence.

What do buyers need to know before they compare sites?

Start with the operating mechanics, not the marketing label. “Demand response capable” can cover very different arrangements. A buyer needs a written explanation of:

• the party that can call an event and the contractual trigger;
• whether participation is voluntary, compensated, mandatory, or tied to a particular service tier;
• the notice process, event duration, and any limits on frequency;
• the amount of contracted load that may be affected;
• what happens if the tenant cannot reduce consumption;
• how a curtailment interacts with backup generation, batteries, maintenance windows, and uptime commitments.

Those questions are not an attempt to negotiate against the utility from a colocation sales call. They are basic workload planning. A company running batch model training, asynchronous data processing, or a non-production environment may have room to shift work. A company supporting payments, emergency services, or real-time inference serving customer applications may not.

The mistake is treating the customer’s rack load as one thing. In practice, the customer may be able to identify a narrow, schedulable portion of demand while needing the rest to remain untouched. That distinction should show up in the proposal, the capacity schedule, and the operating escalation path.

How should operators present flexible power without scaring buyers off?

Lead with firmness. State exactly what power is committed for the tenant’s critical load, what sits outside that commitment, and what assumptions support each statement. Then explain any flexibility program as a bounded option or obligation, not a vague feature.

A useful sales conversation sounds more like this: “Your production environment is supplied under these terms. If you choose to place interruptible workloads here, this is the notice process, this is the load block involved, and this is how we document the benefit.” It should not sound like: “The grid is tight, but we’re working with the utility.”

Be equally plain about where flexibility occurs. It may be at the utility meter, within a customer’s cage, through site-level load sequencing, or via an energy center that supports the campus. Those are materially different designs and risk profiles.

Behind-the-meter systems are making that distinction more important. Worley recently announced engineering and procurement work for data-center energy centers combining gas turbines, aeroderivative turbines, reciprocating engines, battery storage, and intra-campus distribution (Worley). A buyer does not need an engineering lecture, but they do need to know whether the proposed resilience and flexibility story relies on utility supply, on-site assets, or both.

Honestly, operators should avoid presenting on-site generation as a magic answer. Buyers will ask about fuel arrangements, emissions permissions, maintenance, test procedures, cost allocation, and how the system behaves during a true grid event. Have the people who can answer those questions available before the deal reaches final redlines.

Which buyers will care most?

Enterprise infrastructure teams will usually focus first on operational exposure: can they keep applications running, and who owns the incident process? Their finance and procurement colleagues will then ask whether flexibility changes their rate, commitment, or exposure to pass-through costs.

Brokers and site selectors tend to use the issue differently. They are comparing claims across several markets and want a clean basis for comparison. Give them a concise power-terms sheet that separates utility service, customer obligations, expansion conditions, and resilience architecture. If they have to translate your language into a usable comparison table themselves, you have made your facility harder to recommend.

AI buyers deserve extra care. “AI” is not a load profile. A training deployment, an inference cluster, a model-development environment, and a mixed enterprise platform do not have the same tolerance for interruption or for staged energization. Ask what can move, what cannot, and who can authorize a load reduction. Do that before quoting an attractive rate tied to flexibility.

How does price volatility change the conversation?

Price volatility is one reason buyers are taking these clauses seriously, even when they do not expect to curtail often. The U.S. Energy Information Administration reported a wide spread in PJM’s June on-peak daily prices, and noted that peak demand approached or reached annual highs in several major regions (EIA).

You should not use market volatility as a scare tactic. But you should be ready to explain which energy costs are fixed, indexed, passed through, or affected by a customer’s behavior. A buyer who learns late in diligence that the lowest quoted occupancy cost depends on a demand-response construct will feel misled, even if the contract technically disclosed it.

The stronger approach is to present a base case and a flexible-load case side by side. Show the operational conditions, commercial treatment, and decision owner for each. That gives the customer a choice they can take to their operations and finance teams.

Common questions

Does flexible demand mean my data center power is unreliable?

Not automatically. It means you need to establish which portion of your load is subject to a flexibility arrangement and whether your critical systems remain under firm service terms. Reliability claims should be tested against the actual utility agreement, facility design, and tenant contract.

Can we use demand response for AI workloads?

Sometimes, but workload classification comes first. Schedulable training or non-production processing may be a better fit than latency-sensitive inference or customer-facing applications, and your operations team needs authority over any change.

Should a broker treat flexible power as a negative in a shortlist?

No. Treat it as a comparison factor, much like a phased delivery date or a utility-cost pass-through. A clearly documented program can be useful; an undefined promise should reduce confidence.

When should we raise this with a prospect?

Raise it in the first serious power discussion, before the proposal creates an assumption of fully firm, unconditional service. Waiting until legal review turns a manageable operating question into a trust problem.

Where this leaves you

Flexible demand is becoming part of how capacity gets financed, approved, and used. The operators that win trust will not hide it or oversell it; they will translate it into clear workload, contract, and operating choices. GridReach helps data center and energy companies turn expertise like this into qualified pipeline.