Large-Load Tariffs: The New Price of Data Center Power

For the past two years, the data center industry’s power story has been about getting in line: multi-year interconnection queues, transformers on backorder, gas turbines sold out through 2030. In 2026, a second constraint showed up that doesn’t care how fast you can build — it shows up after you’ve already won your interconnection slot, in the form of a utility tariff that tells you exactly how much you will pay, for how long, whether you use the power or not. Ohio and Virginia regulators have now approved large-load tariffs that lock data center customers into 12-to-14-year contracts with minimum-payment floors of 85% of reserved capacity, and in Virginia’s case, collateral requirements that can run into the hundreds of millions of dollars for a single campus. Wood Mackenzie expects data center-specific tariffs to be in force in at least a dozen states by the end of 2026. For anyone negotiating a power agreement right now, this is no longer a regulatory footnote — it’s the term sheet.

Ohio wrote the template: 85% minimum take, 12 years, and a queue that shrank by more than half

AEP Ohio’s data center tariff, approved by the Public Utilities Commission of Ohio, became the reference case other utilities are now copying. The numbers explain why. AEP Ohio’s data center load grew from roughly 100 MW in 2020 to about 600 MW in 2024, with 5 GW projected by 2030 — but the utility had also received over 30 GW of interconnection requests, more than three times Ohio’s own 9.4 GW system peak. AEP froze new service requests in March 2023 while regulators worked out how to avoid building grid infrastructure for load that might never show up.

The tariff that resulted requires data centers drawing more than 25 MW to commit to at least 85% of their contracted capacity for 12 years — a four-year ramp from 50% to 90% minimum payment, followed by an eight-year firm commitment — with a three-year exit notice period or an exit fee equal to three years of energy usage if a customer leaves early. The effect was immediate: once the tariff took hold, AEP’s projected data center demand dropped from roughly 30 GW to about 13 GW, as speculative projects that couldn’t commit to those terms dropped out of the queue. That’s the mechanism regulators wanted — separating real, financeable load from capacity reservations nobody intended to build — but it’s also a filter buyers now have to clear. Duke Energy Ohio filed a competing large-load tariff with the PUCO in late 2025, and the Ohio Manufacturers’ Association is challenging the AEP tariff in the Ohio Supreme Court, arguing it sets a precedent for discriminatory rate classes against energy-intensive customers generally. The legal outcome is still unresolved, which matters for anyone assuming the current terms are final.

Virginia went further: collateral as a stranded-cost filter

Dominion Energy’s new GS-5 rate class, approved by Virginia’s State Corporation Commission in late 2025 and taking effect January 1, 2027, applies to data centers drawing 25 MW or more — which the SCC estimates covers most of the roughly 450 data centers Dominion already serves in the state. The headline term is collateral: $1.5 million per megawatt of booked capacity, meaning a 250 MW campus has to post roughly $375 million before it gets connected. Dominion originally proposed that figure; data center interests pushed for $450,000 per MW instead, and the Commission kept Dominion’s higher number. On top of the collateral, GS-5 customers pay for at least 85% of reserved transmission and distribution capacity and 60% of reserved generation capacity over a 14-year term, regardless of actual usage.

The justification traces back to a 2024 Joint Legislative Audit and Review Commission (JLARC) study, cited repeatedly in the SCC proceeding, which estimated that unconstrained data center growth could add up to $18 billion to Virginia’s generation and transmission costs by 2040 if the buildout outpaces what data centers actually pay for. Notably, that same study found no evidence that cost-shifting to residential ratepayers had already happened under the old rate structure — the tariff is explicitly forward-looking, aimed at the risk of stranded infrastructure if a developer reserves capacity and then walks away or never ramps to the load it projected. A Southern Environmental Law Center attorney involved in the case described the collateral requirement plainly: it’s designed to filter out speculative load before it ever reaches construction.

Why regulators are doing this now

The logic is a version of standard utility cost-causation principle — the customer that causes a cost should pay for it — applied to a category of customer that is unusually large, unusually fast-growing, and unusually likely to over-reserve. A single hyperscale campus can require as much new transmission and substation capacity as a small city, built on a timeline measured in single-digit years. If that customer reserves 300 MW, gets the utility to build for it, and then only ever draws 150 MW — because a project gets delayed, a lease falls through, or a company’s compute plans change — the utility has still sunk the capital, and under a conventional rate structure, residential and commercial ratepayers would absorb the difference through higher general rates. Large-load tariffs push that risk back onto the customer that created it, via minimum-take floors, long contract terms, and in Virginia’s case, cash collateral.

This sits one layer below the federal interconnection reforms that FERC adopted for large-load generator interconnection (covered in our look at FERC’s new large-load interconnection rules) — FERC governs how a load gets in the queue and studied; state commissions like the PUCO and the Virginia SCC govern what that load pays once it’s connected. Both layers are tightening in the same direction at the same time, which is why the practical lead time for a data center with firm, bankable power now runs years longer than the physical construction timeline alone would suggest.

What this actually changes in your underwriting

For a buyer evaluating a self-build, build-to-suit, or large colocation commitment tied to new utility service, these tariffs change the deal in three concrete ways. First, the effective cost of capacity is no longer just the quoted rate per kWh — it’s that rate multiplied by a contractual floor (85% in both Ohio and Virginia) applied over a decade or more, so underused capacity is a real, modelable liability rather than a rounding error. Second, exit risk has a price tag: Ohio’s three-years’-energy exit fee and Virginia’s up-front collateral both mean that changing your mind about a site, or needing to scale down because an anchor tenant’s plans shift, carries a cost that should sit in the deal model from day one, not get discovered during negotiation. Third, speculative capacity reservations — holding a grid slot “just in case” while you finalize financing or a tenant — are now expensive enough that utilities expect them to disappear, which is exactly what happened to AEP Ohio’s queue. If your project can’t commit to the minimum-take terms, you’re effectively telling the utility you shouldn’t have the interconnection slot, and it will likely go to someone who can.

There’s a reasonable counter-argument worth weighing before treating this as purely bad news for buyers: tighter tariffs also mean a cleaner, more credible queue. Projects that survive AEP Ohio’s 85%/12-year test or Dominion’s collateral requirement are, by construction, backed by real capital and real demand — which is useful information if you’re trying to assess whether a colocation provider’s “committed pipeline” is actually going to get built, or whether a competitor’s announced campus is likely to materialize on schedule. The tariffs are also not uniform or finished: the Ohio Manufacturers’ Association’s Ohio Supreme Court challenge could still reshape or strike down the AEP structure, and most states haven’t adopted anything like it yet — Wood Mackenzie’s 12-state estimate for 2026 still leaves the majority of U.S. jurisdictions without a comparable framework, which is itself a variable worth tracking by site.

For anyone negotiating power capacity right now, the practical move is to treat the utility tariff with the same diligence as the lease itself: model the minimum-take floor against your realistic ramp schedule, price the exit fee or collateral into your downside case, and confirm which rate class and contract term actually apply in the specific utility territory you’re targeting — because Ohio, Virginia, and whichever state is next are not writing the same rules. Our Data Center Buyers’ Toolkit walks through exactly this kind of due diligence for capacity contracts, and is worth a look before you sign anything tied to a new large-load tariff.

Written by

Raajeev Ratra

Data Center Infrastructure Expert | 15+ Years in DC Design, Operations & Project Management

Raajeev is a seasoned data center professional with hands-on experience in hyperscale facilities, colocation design, power & cooling infrastructure, and global DC operations. He shares practical insights to help engineers and IT leaders build better infrastructure.

Connect on LinkedIn →