With increasing evidence that data centers are driving up utility costs and are maybe propping up an otherwise flagging economy, an emerging tool called “large load tariffs” may just be the thing that saves ratepayers if this whole AI bubble pops. They’re the latest rage at utility regulatory commissions - and as this goes to press, Evergy Kansas just signed on the line for the hottest new large load tariff.
Before we get started, let’s get a little clarity. What is a large load tariff anyways? In short, a large load tariff is an electric utility rate that puts specific financial requirements on really big electric customers. ‘Large load;’ in this case just means ‘really big customers’ who demand a lot of electricity, and a tariff is a rate schedule, or the terms that describe how much a customer is charged and their service conditions.
Your electric bill is determined by a rate schedule, or tariff. A large load tariff is just a special version of that rate schedule for really big customers, like data centers. Think of a large load tariff as a prenup between the utility and data center designed to cut back on speculation.
What’s in a Forecast Anyways?
You might think that a utility faced with an onslaught of data centers might be really enthusiastic about the potential for new revenue and development - after all, who doesn’t want to be the next Levi Strauss, selling pants for the 49ers during the gold rush? But it’s a little different for electric utilities in the US.
Faced with an obligation to serve, utilities are finding themselves extended, looking for new generation capacity and energy, building new transmission lines for data centers, and retooling their systems to make sure that they can serve the thousands of megawatts of data centers that say they’re coming online. And for a utility that wouldn’t be a problem but for two really big issues:
Generation isn’t getting online: A ton of the infrastructure that’s been trying to get online – particularly like wind and storage – has been held up, cancelled, or outright blocked by the Trump administration, and the administration’s preferred fallback to gas is pretty deeply limited by turbine availability (not to mention pollution). That leaves utilities scrambling to figure out how they’re going to meet requirements.
Utilities have to guess if data centers are actually coming: In a previous article, I talked about the 700 gigawatts of data centers that have been knocking at the doors of utilities, clamoring for interconnection. There’s little doubt that our data center needs don’t surpass the total electricity consumption of the entire United States, and therefore most of those requests are likely speculators or repeat asks from the same big companies - like Amazon, Google, Meta, and Microsoft (what are sometimes termed ‘hyperscalers’).
The good news is that most utilities look at that queue with a fair bit of skepticism, but when you have as much as a 7:1 ratio between pure speculation and forecast demand, even a little bit of error can go a long way. Indeed, with average data center campuses now up at around 300 MW, it doesn’t take a lot of error to make a utility look for new infrastructure to serve just a single customer.
A few utilities now explicitly discount the capacity demands from data centers that are asking for interconnection. For example, Nevada Energy knocks 85% of demand off the top of data centers that haven’t signed formal interconnection agreements yet in its forecast, and Duke in the Carolinas claims that it only accounts for data centers that have signed a letter of intent, putting some money on the table – but they’re also pursuing gigawatts of new gas generation with a pretty thin number of agreements in place already according to a spring 2025 report. And many other utilities don’t really disclose how they conduct their forecasting for these giant new customers, or how they’re protecting ratepayers’ interests. So what can we do to protect the interests of utilities and every day ratepayers given this onslaught?
Enter the Large Load Tariff
Every electric customer is served under some form of tariff, so the concept of having a tariff or rate schedule for big customers isn’t particularly new. But what is new are the protections that are starting to crop up for the utility, and hopefully other customers, with the rise of the large load tariff. Large load tariffs are like a public declaration by utilities that if you want it, you’re gonna have to put a ring on it. They’re meant to bind really large customers to the utility so that the customer doesn’t ghost the utility - or other ratepayers.
And why would a large customer slip out in the dead of night? Well, maybe they don’t actually have customers, or a viable business model. See that earlier note about 700 GW of data center developers lurking around, looking to hook up… to the electricity grid. But utilities are getting savvy: rather than just trust that those two guys in a van actually represent billions of dollars of real investment, they’re demanding that prospective large customers sign publicly-vetted prenups in the form of large load tariffs.
Arguably, the first – or at least most prominent – large load tariff was proposed by Ohio Power Company (OPCo), in May of 2024. OPCo’s proposed tariff (since approved) put forward several novel requirements for big customers:
It's for the big kids. Unlike some “incentive rates” that are meant to attract large load customers, the large load tariff required that any customer above a certain size threshold (25 MW in OPCo’s case) would have to take service with the terms of the tariff. The threshold was set in a way that excluded most existing industrial customers, but wrapped in large data centers.
Until death do us part. Or 15 years. The tariff required that large load customers would be bound to the terms of the tariff for eight or more years, with the logic that the utility would have to build long-run infrastructure, and didn’t want to be in the position of not having a customer to pay for it. The tariff doubled down on the seriousness of the long-run contract by setting a high bar to exit the tariff early, with exit fees that helped make sure that the data center customer would stick around, and even put in place restrictions on how much a customer could change its contracted capacity.
Let’s not call it alimony. The tariff put in place a minimum bill, based on the amount of capacity that a customer claimed that they would need; so if a customer came to OPCo claiming that it would need 100 MW, the tariff would require that customer to pay for at least 85 MW of capacity every month, irrespective of if they used it or not.
Deadbeats not welcome. Finally, the tariff put in place a stiff collateral requirement, requiring that less credit-worthy customers put down a credit guarantee… just in case they happened to go bankrupt.
The net effect of the large load tariff is that it forces even well-financed customers to get serious about their requirements and monetize their commitment. And for truly speculative customers, the real estate dealers working out of their apartment in Florida, it would likely be really difficult to get the financing needed to back that tariff (and collateral) without a bulletproof business plan. As a result, the utility anticipated that the tariff would help it get control of its data center forecast by putting in place a strong market signal.
Like any innovation, there was opposition from the incumbents. The big kids, like the Data Center Coalition, a trade group representing data centers, complained vociferously, and even put out their own competing “settlement”, without the utility on board (weird flex, but okay). But eventually the tariff was finalized with terms similar to those offered by OPCo – and it kicked off a regulatory firestorm that’s largely escaped the public’s eye.
Shortly after AEP filed its Ohio large load tariff, it followed with a similar measure in Indiana, and then utilizes were off to the races, as shown by this great tracker from the Smart Electric Power Alliance. As this article is going to press, there are large load tariffs under consideration in Arizona (APS), Florida (Duke), Kansas (Evergy), Kentucky (KU/LGE), Michigan (Consumers), Missouri (Evergy and Ameren), Texas (SWEPCO), Virginia (Dominion, APCo, and Mecklenburg, Shenandoah, and Rappahannock cooperatives), and Wisconsin (WEPCo) – and while they all differ, they have in common an intention to protect the utility and other customers from the risks of data centers.
The Economy’s Pocket(book) Protector?
There’s a lot to say about what’s getting interesting in large load tariffs, but for that you’ll just have to stay tuned for the next article. But why am I doting on ratemaking? I’m glad you asked.
Large load tariffs – if designed well – play a critical role in holding data centers (and other large loads) accountable for their forecast energy consumption, and take utilities out of the business of being the arbiters of what’s real and what’s not – and that makes all the difference to a utility’s future health, the affordability and reliability of our grid, and even our own health.
Over the last two years, we’ve seen more utilities than we can shake a stick at pour their hearts into proposals for new gas plants to serve data centers (like Dominion, Duke, Evergy, Georgia Power, American Electric Power, Entergy Arkansas, Wisconsin Electric… the list goes on), and in many of these cases there is not much clarity about the level of demand that’s actually sitting behind those proposals. And this Administration has been particularly enthusiastic using the perception of unbounded data centers to create a false sense of emergency, and use that to prop up aging coal plants.
As I talked about in Fools Gold, utilities have been facing an extraordinary amount of pressure from both real, financed technology companies and somewhat more fictional proposals from real estate companies looking to get a piece of the action. These more speculative companies are looking to either just buy up the rights for land, fiber, and power, build and flip data centers, or are just plans on paper. The problem is that utilities face this “obligation to serve,” meaning that utilities have to plan, and build, for the demand that’s going to be coming around the corner, irrespective of where it comes from. So if you’re a utility facing 50 GW of data center requests, should you plan on snapping up every last turbine?
If utilities overextend themselves to build electricity infrastructure for speculative data center customers, and the bubble bursts, utilities will not waste a moment asking other ratepayers (i.e. you and me) to foot the bill. With skyrocketing electricity prices, utilities just can’t afford to hold that risk - at least not without some serious ratepayer - and policymaker - rebellion. If it turns out that really big AI is more hype than monetizable value, we could see a lot of that apparent support for utility infrastructure quietly slip away in the night. And unlike other utility risks - like bad weather, natural disasters, cost overruns at a nuclear power plant, or a steel mill closing, the large-scale reduction of a data center class of customers would hit numerous utilities hard, simultaneously.
Large load tariffs serve a critical role here. They’re designed with two outcomes in mind. First, if that AI boom goes bust, or even just retracts a bit, that the utility still has a paying customer (or a bank, anyways) who will make it whole. Second, if a customer can’t make the case to a credit-worthy financial institution that it has a reasonable chance of making it even if the rocket fuel runs out, then it’s not going to be able to meet the requirements of the tariff. No (reasonable) bank would put itself on the line to finance 15 years of utility bills if it didn’t have really good evidence that the customer has a strong business case. We’re already seeing evidence that Ohio Power Company’s large load tariff is doing what it’s supposed to do, when the utility announced that its speculative ‘pipeline’ had been slashed in half.
Large load tariffs force companies and utilities to take a deep breath and sort the wheat from the chaff, and forces financial institutions to try to nail down what’s real and what’s not. And given that the data center boom now seems to be propping up a good portion of the US economy, maybe some guardrails on speculation are the safety net that we all need.
Correction: The prior citation in this article to an EEI assessment of utility capital expenditures should have noted that expenditures are not exclusively to serve data center load.