As data center energy demand continues to explode, utilities and data center operators have been forced to contend with how to integrate these uniquely large demands into existing tariff and rate frameworks. Many have found that innovative tariff structures are required to meet energy needs while ensuring reliable interconnection to the grid.
The stakes are significant—potentially hundreds of millions of dollars in annual costs. Optimal solutions will vary from situation to situation, but several considerations are likely to become nearly universal.
Operator vs. utility objectives
Each data center operator doesn’t necessarily share the concerns of the next, but two are common: minimizing costs and adhering to sustainability goals. Large tech companies navigating the AI development race have tended to place more weight on sustainability concerns, while others have prioritized keeping costs in check. Most have at least some appreciation for both.
Utilities have focused on protecting bottom lines and cost allocation. Opportunity to earn a return on investments and the significant efforts required to serve large loads is at top of mind, as is isolating other ratepayers from the potential for incremental costs brought by new loads. Utilities also have limited risk tolerances given the cap on ROEs that can be earned to compensate for additional risk.
These disparate objectives don’t always clash. Disagreements have been focused on the terms of service rather than cost-of-service principles that have guided ratemaking for years. AWS has challenged AEP Ohio’s proposal that it be subject to charges for a minimum of 90% of its contracted demand. AWS’ challenge raises concerns with an alleged violation of cost-causation principles due to its sole application to data center loads, while other tariff schedules, like general service, are subject to a 60% minimum.
On the sustainability front, operators and utilities are seemingly more aligned. Operators with sites outside organized markets have sought to approximate the market environment by securing access to supply options other than default service. Several utilities already have some version of programs that permit access to renewable resources, such as Duke Energy’s Green Source Advantage (“GSA”) and Georgia Power’s Clean and Renewable Energy Subscription (“CARES”). Though such programs can support the development of renewable capacity, contract durations and available capacity limitations present challenges to large loads. These programs also remove utilities’ option to earn a return on invested capital in generating resources.
The rise of dedicated resources
Rather than trying to fit carbon-free supply needs into existing market access models, some data center operators and utilities have begun exploring alternatives. Google and NV Energy recently announced an agreement to supply a Nevada data center with a dedicated geothermal resource developed by Fervo Energy. Google will take delivery service under an existing rate schedule, receive capacity and energy credits for avoided production costs, and layer costs for the dedicated resource on top of existing schedule rates through the newly created Clean Transition Tariff (“CTT”).
Despite dedicated resources having gained traction, challenges remain. Those considering less mature technologies may need to proceed cautiously as utilities balance regulatory oversight and stakeholder obligations. Still, evidence of a shift is mounting—Google has noted it considers small modular nuclear reactors (“SMRs”) as a potential energy source and Oracle anticipates three SMRs to supply one gigawatt of Oracle data center load.
Existing tariffs and potential changes
Utilities have responded to increased risk of shifting costs with take-or-pay provisions, such as AEP’s proposal for minimum charges equal to 90% of contract capacity. Others, like Duke Energy, have done the same. Duke recently secured approval in North Carolina for a requirement that loads exceeding 100 MW enter a minimum payment contract. In another example, Entergy inked a 20-year contract with Meta for a data center in Louisiana, which includes minimum payments. Though some operators have acquiesced, some such provisions have come with broad pushback—AWS’ arguments against AEP’s proposal were joined by Google, Microsoft, and Meta.
The need for sensible risk sharing
The strong opposition to take-or-pay provisions suggests the methods used to contend with risk sharing may yet change. Regardless of the mechanisms used, collaborative approaches toward risk sharing must balance the interests of all parties by fairly distributing financial burdens and potential rewards.
Striking a proper balance will be challenging, but several mechanisms may prove worth exploring:
|
→ Financing contributions or loan guarantees for dedicated resources – large operators with cash-rich balance sheets may fund reduction of utility risk, easing constraints. |
|
→ Joint venture ownership structures – permitting operator investments without operational responsibilities in facilities, in exchange for reduced rates. |
|
→ Abandoned plant guarantees – guarantees for expended capital on projects that may never be placed in service, can reduce risk for novel technologies (e.g., small modular nuclear reactors) in particular. |
|
→ Tranche-based diversification – assigning multiple large loads to individual resources can diversify operator risk, but industry risk will remain if all loads belong to data centers. |
Adaptable and scalable solutions
The service terms defined by this discussion will best serve the industry by effectively contemplating future loads. Tariffs resulting from these discussions can be structured to be adaptable and scalable.
Adaptable tariffs will be useful for a wide range of interconnections, with terms that are not overly prescriptive. A balance must be struck between sufficiently defining terms of service and permitting flexibility across different sites and operators with varying needs.
Similarly, tariffs should be scalable. Constraints will always exist for certain attributes of service, but within these constraints, tariffs should scale with incremental customers and loads. When constraints are redefined, tariffs should continue scaling within the new parameters.
