How Xcel's Proposed Large Load Tariff Changes Data Center Site Selection

A Colorado data center site needs a power plan that connects the requested capacity to the project's actual development schedule. Xcel's proposed large-load tariff makes that connection more consequential.

The proposal would generally apply to new or expanded loads of at least 50 MW. Xcel also says some new customers between 20 and 50 MW could fall within its scope. The utility proposes long-term commitments and responsibility for the infrastructure needed to serve the customer. Xcel's April 2 announcement sets out the framework.

As of September 14, the proceeding remains under review. For developers, the immediate task is to test a site's economics against the proposed terms and preserve enough detail to update the model when the commission acts.

I would begin with three numbers: the power capacity the project contracts for, the demand it expects to reach, and the energy it expects to consume. Each should have a date attached.

The proposed commitment reaches well beyond opening day

Xcel's first-quarter investor presentation summarizes the Colorado proposal as including a 15-year minimum term, an 80% minimum demand charge, credit requirements, and 24 months' notice for capacity reduction or service termination. The presentation also describes an exit fee tied to the remaining contract term. Slide 37 contains the Colorado summary.

Those provisions connect the early site decision to a much longer operating plan.

A developer should be able to explain when the first building starts using power, when later buildings arrive, and how the commercial commitments follow that schedule. The model should also show what happens if a phase starts later than planned.

For example, a project might reach its first operating milestone on time while tenant deployment takes another year. Another project might have tenant demand ready while an electrical upgrade remains on the critical path. Those scenarios put different pressures on the same development budget.

The relevant question is when each payment obligation begins under the actual agreement. That date belongs beside the construction schedule and the expected revenue schedule.

Land diligence should support this analysis early. A site that requires an additional parcel for electrical equipment, a new access agreement, or a separate approval for expansion needs those dependencies reflected in its milestones.

A simple stress test for the 80% demand floor

Consider an illustrative project with 100 MW of contracted demand. Apply an 80% minimum billing-demand assumption during a period when that floor is fully in effect. For this simplified test, billed demand is the greater of measured demand and 80 MW.

This is an original sensitivity calculation using the proposed 80% concept. Actual billing will depend on the approved tariff, the service agreement, ramp provisions, and other applicable charges.

The table shows why the timing of contracted capacity matters. At 40 MW of measured demand, the modeled billing quantity is twice the measured quantity. At 60 MW, the modeled billing quantity is one-third higher.

To express the effect in dollars, every $1 per kW-month of an applicable demand rate would produce an $80,000 monthly charge on 80 MW. The same rate applied to 40 MW would produce $40,000. The difference is $40,000 per month for each $1 per kW-month of demand rate.

That calculation uses a unit rate so readers can insert the rate relevant to their own model. It measures the demand-charge component under the stated assumptions. Energy charges and other bill components require separate inputs. The Energy Information Administration explains the power and energy units used in these calculations.

Under an illustrative 80% billing-demand floor, a 100 MW commitment creates an 80 MW minimum billing quantity. If measured demand is 40 MW, that minimum is twice the measured demand. For a data center developer, the practical question is when the contractual floor begins and how the load ramp aligns with it. The service agreement and approved tariff determine the actual obligation.

This is the kind of sensitivity I would run before assigning a value to a property's ultimate power capacity.

Load factor describes the operating pattern

The Colorado Energy Office has reportedly proposed a 75% to 85% load-factor screen in defining qualifying large customers. Denver has reportedly proposed aggregating commonly owned or functionally integrated sites when applying the tariff threshold. The September 9 reporting on intervenor testimony identifies both positions as proposals in the proceeding.

Load factor is average load divided by peak load over a specified period. EIA's definition provides a clear basis for the calculation.

A facility with a 50 MW peak and a 75% annual load factor averages 37.5 MW over the year. Using 8,760 hours, that equals 328,500 MWh. At an 85% load factor, the same peak corresponds to 42.5 MW on average and 372,300 MWh annually.

The difference is 43,800 MWh per year. That is a meaningful change in the energy requirement even though the peak remains 50 MW.

Load factor and contract utilization measure different aspects of a data center's power use. Load factor compares average demand with peak demand during a defined period. Contract utilization measures use of committed capacity. A facility drawing a steady 40 MW against a 100 MW commitment would have a 100% load factor and 40% contract utilization over that period.

That example is worth keeping in a development model. A project can operate steadily while using only part of its contracted capacity. The model needs both measurements to explain its operating pattern and its commercial exposure.

For an actual project, use the measurement interval and definitions established by the applicable tariff. Then make sure the engineering load profile and the commercial model use the same assumptions.

Evaluate the whole campus when testing eligibility

Denver's reported aggregation proposal makes the relationship among buildings, meters, parcels, and operating entities a relevant modeling input.

A campus review should show common ownership and shared infrastructure. It should identify which buildings share access, cooling systems, electrical facilities, or a common development schedule. Those facts help the team evaluate how a final tariff might apply.

For a simple illustration, three related 30 MW buildings represent 90 MW of combined planned load. The project team should test the campus economics at that combined scale while the eligibility rules remain under review.

The real estate file should make those relationships easy to inspect. A parcel map can show legal boundaries, while a campus plan can show how the operation works across them. Both belong in the same diligence package.

This also improves conversations with potential tenants or buyers. A buyer evaluating one building needs to understand any obligations that arise from shared infrastructure or the broader campus arrangement.

A useful site summary would therefore include the proposed first phase, the ultimate campus load, the related entities, and the basis used to assess tariff eligibility. Each item should be updated as the commercial structure becomes more definite.

Compare sites using the same load schedule

A fair site comparison starts with the same operating assumptions. Give each candidate the same initial demand, the same ramp, and the same target service dates. Then identify how the utility path changes the result.

I would compare the initial infrastructure contribution, recurring demand charges, energy charges, financial security, and the cost of a schedule delay. I would also record which figures come from an agreement, a utility estimate, or a modeling assumption.

That last distinction helps the team choose its next expenditure. If the largest uncertainty is the cost of an electrical upgrade, a utility study may be the next useful step. If a proposed substation requires a separate land approval, the planning path may need attention first.

For a phased project, run at least two capacity schedules. One can match the developer's preferred buildout. Another can delay the later phases while preserving the first operating building. The difference shows how sensitive the project is to the timing of growth.

The demand-floor example makes the exercise concrete. A 50 MW commitment with an 80% floor creates a 40 MW minimum quantity. A 100 MW commitment creates an 80 MW minimum. Whether either structure is available depends on the utility's terms and the project's needs.

The purpose of the comparison is to understand the value of a realistic phase plan. The team can then discuss that plan with the utility using specific dates and quantities.

Keep the utility relationship visible on the map

The serving utility and the transmission provider may have different roles. The commission's June order in this proceeding describes United Power's dependence on Public Service Company's transmission resources for some large-load service arrangements. The order's discussion of United Power shows why that relationship belongs in a site assessment.

For a parcel, the power record should identify who provides retail service, which infrastructure requires study, and which party is responsible for each agreement. It should also show the current status of the relevant tariff or negotiated service terms.

A utility boundary can therefore change the questions a developer needs to answer. Two neighboring properties may move through different commercial and technical processes even when they sit near the same transmission corridor.

The map should help a reviewer see those differences. Clicking on a parcel should reveal the requested load, the proposed service date, the serving utility, and the documents supporting the power path.

That makes the map useful to the people who have to act on it. A broker can explain the current evidence. An engineer can identify the next study. A developer can see which assumptions are ready for a financial model.

Update the model when the commission acts

The remaining proceeding can change thresholds, contract terms, billing rules, and the treatment of related sites. The sensible response is to keep those inputs editable and preserve the version used for each decision.

I would save a dated model before a major utility filing or commission decision. After the new terms are available, rerun the same project assumptions and identify which sites or phase plans change materially.

That history can become useful research of its own. It shows how a regulatory change affects a consistent set of hypothetical projects, with the calculations available for review.

For now, the proposed tariff points to a clear priority: size the initial commitment carefully, model the actual load ramp, and connect both to the land and infrastructure plan. A credible power schedule should be part of the site's value proposition from the beginning.