The Interruptible Campus: Data Centers Trade Uptime for Queue Position
Photo: DeVos Max, CC BY-SA 4.0, via Wikimedia Commons
The data center boom has collided with a grid that was not planned for it. In several US markets, large loads are discovering that the constraint is not land, capital, or even chips. It is the ability to connect. As of November 2026, a new negotiating posture is emerging: data center operators are offering to be curtailed, throttled, or shifted in exchange for faster service and more favorable rates. The bargain is attractive on paper. It is also fragile.
What a Flexible-Load Agreement Actually Says
The phrase flexible load can obscure a highly technical contract. A useful agreement specifies the conditions under which the utility may reduce or interrupt service, the notice the operator will receive, the maximum duration and frequency of events, and the compensation or rate credit the operator receives. It also defines what counts as an event: a grid emergency, a capacity shortfall, a locational constraint, or an economic signal.
The operational details matter as much as the legal language. Can the facility ramp down gradually, or must it drop load in blocks? Does the operator retain control of its own equipment, or does the utility dispatch a signal directly? What telemetry is required? How are penalties assessed if the operator fails to curtail? A flexible-load agreement that is vague on these points is not a reliability tool. It is a dispute waiting to happen.
Tariff Experiments Utilities Are Running
Utilities and state regulators are testing several structures. One is an interruptible large-load rider, in which the customer receives a discounted rate in exchange for accepting curtailment up to a specified number of hours per year. Another is a capacity reservation tariff, where the customer pays for a defined level of firm service and can subscribe to additional interruptible capacity at a lower price. A third is a dynamic pricing arrangement tied to wholesale market conditions, sometimes paired with on-site generation or storage.
Each structure allocates risk differently. A simple interruptible rider shifts reliability risk to the customer but may not give the utility enough control during a localized constraint. A capacity reservation tariff can provide clearer cost signals, but it may be complex to administer. Dynamic pricing can align incentives with market conditions, but it requires sophisticated metering and may expose the customer to volatile costs. The common thread is that utilities are trying to convert a blunt interconnection request into a more granular service product.
The Counterparty Risk Nobody Wants to Price
The central risk is that the data center operator may change its mind. A hyperscaler may decide that a particular campus supports latency-sensitive workloads that cannot be interrupted. A colocation provider may sell the facility to a customer with different service requirements. A tenant may insist on contractual uptime guarantees that conflict with the utility's curtailment rights. Even a willing operator may resist curtailment during peak revenue periods, creating tension between the contract and commercial reality.
Utilities face a second risk: stranded network upgrades. If a large load receives a fast-track interconnection on the basis of flexible operation, the utility may size substations, transformers, and lines for a higher level of service than the customer ultimately uses. If the customer later reduces its load or exits, those costs may be borne by remaining ratepayers. Regulators are beginning to ask whether flexible-load tariffs should include exit fees, minimum take obligations, or collateral requirements.
Designing a Durable Bargain
A durable agreement must be designed for change. That means clear curtailment priorities, measurable performance obligations, and compensation that reflects the value of flexibility rather than a one-time interconnection concession. It means diligence on the operator's business model, not just its credit rating. And it means utility planning that treats flexible load as a resource with uncertain availability, not as firm capacity.
For data center operators, the trade-off is equally real. Accepting curtailment can accelerate time to power and reduce costs, but it can also compromise service-level commitments to tenants. Operators should model curtailment scenarios against their revenue contracts and be transparent with customers about the limits of interruptibility.
The emerging bargain is not a solution to the interconnection backlog. It is a way to extract more value from existing capacity while the grid catches up. Handled well, it can benefit both sides. Handled carelessly, it becomes a cost allocation problem that regulators, utilities, and customers will litigate for years.
