As policymakers debate federal permitting reform, AI regulation and industrial policy, a quieter but increasingly consequential battle over artificial intelligence infrastructure is unfolding before state public utility commissions.
According to a new analysis published by Tech Policy Press, state utility regulators are rapidly becoming some of the most important decision-makers shaping where AI data centers are built, how quickly they can be connected to the electric grid and, perhaps most importantly, who pays for the infrastructure needed to support them.
The analysis argues that many of the most contentious issues surrounding AI infrastructure development are not fundamentally AI policy questions at all. Instead, they are longstanding utility regulation questions involving cost allocation, ratepayer protection, investment prudence and infrastructure planning.
As of early 2026, 27 states had pending legislation addressing how new “large-load” customers, a category that includes most hyperscale and AI-focused data centers, should be evaluated, priced and connected to electric systems.
The growing role of state regulators comes as communities, consumer advocates and lawmakers increasingly push back against the cost and energy demands associated with AI infrastructure. Critics have raised concerns that data centers could drive higher electricity rates for households and businesses if utilities are allowed to socialize the costs of new power plants, transmission lines and substations across the broader customer base.
The Tech Policy Press analysis suggests that existing utility regulatory frameworks may provide tools for addressing precisely those concerns.
At the center of the debate is the principle of “cost causation,” a foundational concept in utility regulation that seeks to ensure that customers causing infrastructure investments bear the costs associated with them. Traditionally, utility costs are spread across ratepayers because load growth occurs gradually across many customers. AI data centers challenge that model because a single facility can require infrastructure investments comparable to those needed for entire customer classes.
Idaho’s recently enacted House Bill 911 illustrates how states are adapting traditional utility principles to AI-era demands. The law, which takes effect in July 2026, requires any new load exceeding 50 megawatts to receive service under a utility commission-approved contract. Those agreements must satisfy a statutory “no harm” test and demonstrate that the customer will fund its share of generation, transmission, substation and distribution investments needed to serve the facility.
Related: FERC Orders Grid Operators to Fast-Track Data Center Connections
Such requirements directly address one of the most politically sensitive aspects of the data center debate: whether ordinary ratepayers should subsidize infrastructure built primarily for AI developers and large technology companies.
The analysis identifies three additional areas where state regulators are establishing rules that could significantly influence future AI development.
One is prudency review, the traditional utility standard used to determine whether investments were reasonable when made. Because forecasts for AI-related electricity demand vary widely, regulators must decide how aggressively utilities can commit to new generation and transmission projects based on anticipated data center growth. Stricter standards could require firmer commitments from developers before utilities build infrastructure, while more permissive approaches could accelerate deployment but increase financial risks for consumers.
Another emerging issue involves stranded-cost risk. Utilities that build facilities to serve anticipated data center demand may be left with unrecoverable investments if projects are delayed, downsized or canceled. Several states are responding by requiring developers to provide financial security or other guarantees before utilities commit significant capital. Idaho’s law explicitly requires protections against stranded costs and unrecoverable investments.
The analysis further argues that federal discussions often overlook the complex jurisdictional reality governing AI infrastructure. While the Federal Energy Regulatory Commission oversees wholesale power markets and interstate transmission planning, state commissions retain primary authority over retail rates, resource planning and many generation investments. Major data center projects frequently require approvals at both levels.
For businesses, utilities and policymakers, the implication is increasingly clear. The future of AI infrastructure will be shaped not only by Congress, federal agencies or White House initiatives, but also by state utility proceedings that determine who bears costs, how risks are allocated and what protections exist for consumers.
