Brookings Institution이 AI 데이터센터 비용 영향으로부터 요금 납부자 보호 약속의 이행 강제를 촉구했다.

In less than a year, a consensus has emerged regarding the need—and potential methods—to shield residential and other utility ratepayers from electricity rate increases driven by unprecedented data center demand. Twelve months ago, there was no agreement that a problem even existed. Despite widespread reports of rising electric rates, data center advocates insisted that costs would fall, not rise, as new facilities came online. Today, however, it is widely accepted that non-data-center rates will increase unless checked. A consensus has also formed around specific actions to moderate or prevent those hikes, with the largest AI data companies pledging their commitment to ratepayer protection.

Yet the debate is not over, and more lies ahead as policymakers consider what needs to happen next to translate words of commitment into action.

The problem comprises two core elements. First, without new tariff structures, hundreds of billions of dollars in grid upgrades required to meet explosive data center power demand will be spread across all utility customers, pushing residential and non-data-center rates beyond affordability. Satisfying this demand necessitates costly expansions across generation, transmission, and distribution infrastructure. An ICF report projects that, absent new tariffs, residential electricity rates will rise between 15% and 40% by 2030, with some utilities seeing rates double by 2050. Second, there is the risk of “stranded investment.” What occurs if utilities build new capacity for data center demand that never materializes or vanishes within a few years? Such a scenario could unfold if the AI “bubble” bursts, if soaring electricity costs compel chipmakers and developers to adopt energy-saving efficiencies, or if competitive dynamics realign the market. Market intelligence firm Sightline Climate recently estimated that up to 50% of the 2026 announced pipeline for large data centers may never come to fruition. This aligns with Wood Mackenzie data showing that new data center announcements dropped by half between the third and fourth quarters of 2025.

The emerging consensus on addressing these challenges is embodied in the “Ratepayer Protection Pledge,” announced by President Donald Trump on March 4, 2026. Under the pledge, the hyperscalers driving massive data center development—Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI—committed to:

• Build, bring, or buy new generation resources and cover the cost of all power delivery infrastructure upgrades required for their data center.

• Negotiate separate rate structures with utilities and State governments.

• Pay these rates for the power and related infrastructure brought online to service their data centers, whether they use the electricity or not.

• Coordinate with grid operators to make backup generation resources available, contributing to a more reliable grid and preventing blackouts and power shortages in times of emergency.

To gauge how rapidly the issue has evolved, note that less than six months prior, U.S. Department of Energy Secretary Chris White issued a 16-page directive to the Federal Energy Regulatory Commission (FERC) to accelerate data center grid interconnections without once mentioning the word “ratepayer.”

Though drafted in broad terms, the pledge directly addresses both core challenges. Establishing new tariff structures that isolate the incremental costs of generating and delivering power to new data centers is essential to any ratepayer protection program. Without isolating these expenses and assigning them to developers via separate tariffs, non-data-center bills will inevitably rise. The pledge’s second pillar—the requirement that developers pay for power “whether they use the electricity or not”—is equally critical to shielding ratepayers from stranded investment risks.

Beyond its operational terms, the pledge aimed to reshape the political landscape. Signatories clearly intended to make it politically viable for federal, state, and local officials to champion ratepayer protections while still supporting AI expansion. Leading AI firms have reinforced this public commitment, spearheaded by Microsoft President Brad Smith, who unveiled his company’s five-part “Building Community-First AI Infrastructure” plan earlier this year. All pledge signatories have since echoed these assurances in public statements.

While the pledge and industry commitments mark significant progress toward a new consensus, it remains unclear whether they will curb rapidly escalating public opposition to massive data center construction. A February 2026 Marquette University Law School poll of Wisconsin voters revealed that 70% believe the costs of new data centers outweigh the benefits—a sharp rise from 55% just six months prior. Additionally, 35% agreed that data center development “drives up the cost of electricity for everyone else.” In Virginia, home to “data center alley,” only 35% of voters expressed comfort with a new facility being built in their community in 2026, down from 69% in 2023. The New York Times noted that comedian Charlie Berens described opposition to data centers as “the most bipartisan issue since beer.”

Public concerns regarding AI and data centers extend far beyond ratepayer protection. A recent Gallup poll found that seven in ten Americans oppose building data centers, yet only 15% cited higher utility bills or energy costs as a reason. Instead, 50% cited impacts on local resources, 22% raised quality-of-life concerns, and 16% pointed to environmental issues such as noise, air pollution, or water contamination. Opposition to data centers may well serve as a proxy for broader anxieties about AI. In the same Gallup survey, 14% explicitly opposed data centers due to negative views on AI itself. Meanwhile, a recent Quinnipiac poll found that 80% of Americans are concerned or very concerned about AI. Only 31% trust the U.S. government to regulate AI responsibly, ranking the United States last among nations surveyed by Ipsos in 2025.

Against this backdrop of overwhelming public opposition, legislative efforts at the federal, state, and local levels to impose moratoria on data center construction are increasingly understandable. In March 2026, Sen. Bernie Sanders (I-Vt.) and Rep. Alexandria Ocasio-Cortez (D-N.Y.) introduced legislation to enact a federal moratorium on data center construction, explicitly aiming to pause “the development of AI.” While the bill faces certain defeat in the current Congress, state-level initiatives have gained traction. As of March, eleven states had introduced legislation to temporarily ban data centers. A Maine proposal would have established the nation’s first statewide moratorium, but it was vetoed by outgoing Governor Janet Mills, who subsequently suspended her Senate campaign. Local opposition has proven even more impactful. At least 48 data center projects representing $156 billion in potential investment wer