Last week, Xcel Energy proposed a long-awaited large-load tariff on data centers that want to build in Colorado. The impulse to ensure data centers don’t shift costs onto residential ratepayers is admirable, and this tariff could help in the short term. But if the terms are too restrictive, data centers will bypass Colorado’s grid entirely and ratepayers could end up paying more, not less.
In November 2025, the PUC adopted guiding principles for serving large-load customers and ordering Xcel to file a detailed tariff proposal. The PUC’s principles applied to any customer needing 50 megawatts or more and called for a $250,000 non-refundable study deposit and an early-exit fee equal to 75 percent of all electricity the facility would have used over the full contract life.
In several ways, Xcel’s proposal is better than what the PUC had in mind. The study deposit is $120,000, not $250,000. The exit fee is structured as the sum of remaining minimum monthly bills rather than projected total usage. Data centers would sign a 15-year contract with a security deposit equal to six months of operations and a minimum bill of 80 percent of contracted demand. They’d pay for new generation and transmission built to serve them, bringing total upfront costs to roughly $600,000. And Xcel can’t count data center demand in its load forecast until the customer has signed binding agreements and demonstrated site control.
Data centers, however, are mobile capital. A hyperscale builder will go wherever energy is cheapest and fastest to connect. Residential customers aren’t going to sell their home over a rate increase; they’re inelastic customers next to data centers. A tariff that charges the most price-sensitive customers the highest effective rates will predictably drive them away, either to other states or to behind-the-meter generation.
The experience of American Electric Power in Ohio is instructive. AEP Ohio implemented a tariff in July 2025 that included an 85 percent minimum demand charge, 12 year contracts, and steep exit penalties. AEP’s forecasted large-load demand fell by more than half, from 30 GW to 13 GW, but the capital plan didn’t shrink to match. AEP is still pursuing $72 billion in infrastructure spending originally justified by data center demand projections.
The PUC approved 6,000 MW of new generation for Xcel Energy, down from the 14,000 MW the company requested, but still the largest resource acquisition ever approved in Colorado. If Xcel’s tariff drives away the customers it was designed to charge, ratepayers could still be on the hook for the infrastructure.
There’s a deeper problem with how monopoly utilities set these terms. A data center paying a premium for faster grid connection is making a rational market choice. A monopoly utility forcing customers into 15-year take-or-pay contracts with steep exit penalties, with no alternatives available, is something else. A private business can turn away customers for any reason, but a regulated utility with a government-granted monopoly has an obligation to serve. When that monopoly imposes terms so restrictively that customers are better off building their own power plants, it isn’t fulfilling that obligation.
Colorado’s situation is arguably worse than Ohio’s. Ohio is a partially deregulated market where large customers have some ability to shop for competitive supply through PJM. Colorado has no retail choice whatsoever. If you want grid power in Xcel’s vertically-integrated monopoly territory, you buy it on Xcel’s terms, or you build your own.
The legislature should allow data centers to pursue consumer-regulated electricity, which are privately financed, physically islanded systems serving new large loads that were never on the existing grid. CREUs let private capital take both the risk and the reward. Because the systems are physically separate from the grid, there is zero impact on residential ratepayers. And a market-based alternative will discipline Xcel’s terms more effectively than any regulatory proceeding.
A well-designed tariff keeps data centers on the grid, where their load spreads fixed costs and lowers rates for everyone. Xcel’s proposal could risk driving away the cash cows it is counting on to finance new infrastructure.
