Avoiding the Negative Impacts of Large Electric Loads
by Bob Shively, Enerdynamics President and Lead Facilitator
The rapid expansion of data centers and artificial intelligence applications is creating unprecedented challenges for electric utilities across the United States. A recent Lawrence Berkeley National Laboratory (LBNL) study projects that U.S. data center electricity demand could triple from 2024 to 2028, with 12 states accounting for 84% of recent growth. While these large loads represent important economic development opportunities, they also raise complex issues about how utilities allocate costs, manage investment risks, and maintain grid reliability. The key question is how these issues impact existing utility customers as large loads are added to the grid.
Recent policy initiatives highlight the urgency of addressing these challenges. New York Governor Kathy Hochul's Energize NY Development program and Microsoft's Community-First AI Infrastructure commitment both emphasize a fundamental principle: large-load customers must pay their fair share without burdening residential ratepayers. As utilities and regulators navigate this landscape, three critical issues stand out.
Above: A large data center in the Netherlands
How can utilities be sure that existing customers pay the costs associated with new loads?
The first challenge is ensuring that data centers and other large loads pay appropriately for the electricity infrastructure they require. Traditional rate structures and line extension rules weren't designed for customers that can consume as much power as a small city, and failing to properly allocate these costs risks unfair subsidization by residential and commercial customers. New York's approach requires data centers to either pay rates high enough to cover their full electricity costs or supply their own energy. This "pay your way or power yourself" standard prevents residential customers from subsidizing the massive infrastructure investments these facilities demand. Similarly, Microsoft has committed to working with utilities to set rates that ensure datacenter electricity costs aren't passed on to residential customers.
Utilities are implementing various mechanisms to achieve fair cost allocation. Some require minimum demand charge - for example, requiring customers to pay for at least 80% of their contracted capacity even during periods of lower utilization. Others use marginal pricing methods that directly assign incremental infrastructure costs to the customers causing them. Several utilities now require credit ratings, collateral, or upfront contributions toward construction costs to mitigate financial risks.
These approaches follow a basic principle of utility regulation: customers should pay for the costs they cause. When a data center requires new transmission lines, substations, or generation capacity, those investments should be recovered from that customer rather than spread across all ratepayers.
How can utilities protect themselves from expanding grids for new loads that don’t show up?
The second critical issue involves protecting utilities and ratepayers from the financial consequences of underutilized infrastructure. If a utility builds a new substation or contracts for generation capacity based on projected data center demand that fails to materialize, ratepayers could be left paying for "stranded assets" that serve no productive purpose. Data center electricity demand can be particularly volatile. AI model training may require intensive power during development phases, but less during operation. Cryptocurrency mining operations can shut down rapidly when prices drop. These uncertainties create substantial planning and investment risks.
Utilities are addressing these risks through long-term contracts, typically ranging from 10 to 20 years, that commit large-load customers to specific capacity levels. Many agreements include ramp periods— allowing customers several years to reach full contracted capacity, paired with minimum usage requirements during each phase.
Exit fees provide another layer of protection. If a customer leaves before their contract term ends, they may owe fees covering the remaining minimum charges for unused capacity. Some proposed agreements set exit fees equal to three years of minimum charges. These mechanisms ensure that if demand projections prove overly optimistic, the customer rather than general ratepayers bears the financial consequences.
Several utilities also require customers to resize contracted capacity with substantial advance notice— sometimes 42 months or more — allowing time to adjust procurement strategies and avoid excess capacity costs.
How can utilities manage operational issues associated with large loads?
The third challenge involves maintaining grid reliability when large loads connect or when their demand fluctuates unexpectedly. Data centers can cause voltage and power quality issues, strain local distribution infrastructure, or create resource adequacy problems if insufficient generation capacity is available.
Utility tariffs increasingly specify load factor requirements, defining the ratio between average and peak demand that customers must maintain. Some require minimum load factors of 85%, ensuring relatively steady rather than highly variable consumption patterns that stress the grid.
Many agreements now address behind-the-meter generation, recognizing that data centers often install backup power systems. While backup generation can enhance reliability, utilities are establishing rules about technology types, requiring that equipment remain maintained and available throughout contract terms. Some utilities require access to customer backup generation during grid emergencies, turning a potential reliability risk into a system resource. Others require large loads to interrupt power use when necessary to protect grid reliability.
Microsoft's infrastructure plan includes innovations like closed-loop liquid cooling systems and AI-driven efficiency improvements that reduce both energy consumption and operational stress on utility systems. These technological approaches complement rate design strategies in managing grid impacts.
Moving forward
The convergence of AI expansion and electricity infrastructure planning requires collaboration among utilities, regulators, and large-load customers. Well-designed rate structures can attract economic development while protecting residential customers and maintaining grid reliability. By carefully addressing cost allocation, investment risk, and operational challenges through thoughtful tariff design, the industry can support technological innovation without compromising its fundamental obligation to serve all customers fairly and reliably. Without such care, existing customers may see undesirable impacts.
Read more on Energy Currents
