Backlash Against AI Data Centers in the United States and the Fragmentation of Climate and Energy Regulations: Implications for Middle Powers

Executive Summary

Backlash against AI data centers in the United States is being organized across the ideological spectrum, centered not on climate discourse but on electricity rates, water usage, and safety accidents—a structure that is expected to persist for the next one to two years. While California is tightening regulations by mandating water and power disclosures and cost-sharing, states like Georgia, Texas, and Ohio are absorbing investments through tax incentives, making the balloon effect driven by regulatory arbitrage already a reality. This parallel yet divergent landscape, much like Singapore's proposal for a UN AI safety framework, suggests that international AI norms are likely to be settled in multilateral arenas before individual national legislation, aligning with the broader competition over international norms amid the US-China AI rivalry. In the short term, the South Korean government and businesses must pursue a dual-track strategy: utilizing states with relaxed regulations for site selection, while building proactive compliance capabilities for California-style disclosure and cost-sharing requirements in the long term. Furthermore, they need to detect early signs of convergence between UN climate negotiation channels and AI norm-setting channels to preemptively align domestic data center policies and overseas investment strategies.

I. Issue Analysis

The Clash Between AI Data Center Expansion and Climate Issues: Issue Analysis

1. Background and Progress of the Issue

The flashpoint for the backlash against data centers in the United States was electricity rates, not climate discourse. The Center on Global Energy Policy at Columbia University's School of International and Public Affairs (SIPA) recently analyzed the drivers of rising electricity rates across the US, identifying the surge in power demand from AI and data centers as a key variable [13]. According to a poll cited by TechCrunch, more than 60% of Americans support restricting the construction of new data centers [5]. This figure does not reflect the opinion of any single political faction. The Canadian outlet *The Globe and Mail* commented on this, noting that 'the left and the right have united in their fear and loathing of data centers' [17]. This indicates that opposition is being organized at the local level, regardless of ideological alignments.

According to existing analysis accumulated by the EAI, this structure of backlash 'is being organized on a bipartisan basis around rising electricity rates, water usage, and safety accidents, rather than the climate agenda, and this structure will not fundamentally change over the next one to two years' [3]. The fact that data centers consume vast amounts of cooling water, cause noise and land-use issues, and operate large-scale diesel generators for emergency backup power has directly translated into everyday grievances for local residents. An accident at the Equinix NY2 data center in Secaucus, New Jersey, reported by *Caixin*, symbolically illustrates these concerns. A system failure in a generator's diesel storage tank caused approximately 19,000 to 21,000 liters of diesel fuel to spill into a nearby creek [16]. *Caixin* summarized this along three axes—'pollution, water consumption, and political division'—commenting that data centers, once symbols of high technology, are degenerating into potential sources of petrochemical pollution [16].

This backlash is not confined to the United States; similar local political conflicts erupted earlier in Europe. In the French village of Étrelles, opposition to Google's plans to build a massive data center escalated to the point where an independent mayor resigned, and opposition movements were also organized in Kronstorf, Austria [7]. However, while the European cases remain limited to sporadic resistance at the local political level and discourse on data sovereignty, the United States differs in that the backlash is leading to legislative responses at the state government level.

2. Current Situation

The most distinct recent trend is the tightening of regulations in California. According to reports by *Caixin*, California Governor Gavin Newsom signed seven bills related to data centers on September 21 [18]. The new laws mandate that data center operators disclose their water and electricity usage, and require them to bear the costs of grid upgrades and new clean energy, ensuring that these expenses are not passed on to general residents, including low-income households [18]. Operators must also report water supply, conservation, and drought response plans to local governments and water agencies in advance, and they are no longer eligible for blanket environmental exemptions [18]. EAI analysis defines this as 'regulatory tightening that mandates water and power disclosure and cost-sharing,' and projects that the future path will be a 'parallel yet divergent landscape' contrasted with states like Georgia, Texas, and Ohio, which continue to absorb investments by maintaining tax incentives [3].

This divergence is already manifesting as a shift in jurisdiction by operators. The Carnegie Endowment for International Peace's 2026 California AI Survey reported a case where a Nevada-based utility company recently terminated its contract to supply electricity to residents of South Lake Tahoe, California, choosing instead to focus on constructing data centers within Nevada [15]. EAI interprets this as 'a balloon effect where regulatory tightening in individual states triggers shifts in operators' jurisdictions' [3]. This means that as California-style regulations spread, operators will have greater incentives to relocate their investments to states where tax benefits remain.

*Wired* highlights this phase from a different angle. It points out that despite the growing backlash against data centers, it has not been sufficiently linked to the discourse on addressing the climate crisis [9]. The language of opposition is focused on electricity rates, water, noise, and land use, while connections to greenhouse gas emissions or carbon neutrality targets are treated as secondary. Although Canada's TechInsights raised the structural issue that the growth rate of data center emissions is outstripping the pace of grid decarbonization [2], this awareness has not yet become a primary frame for local grassroots movements in the United States.

In response to this backlash, the industry is taking preemptive measures. According to reports by *Agri-Pulse*, the data center industry has made a series of pledges to reduce water usage, mitigate noise, and avoid contributing to electricity rate hikes [14]. However, experts tracking industry trends worry that these solutions may merely substitute one problem for another [14]. The Brookings Institution also diagnosed that the opacity of land-use agreements, the burden of electricity rates, and environmental and employment concerns are intertwined, creating a structure where it is difficult for the industry, local communities, and governments to reach a consensus [6].

3. Key Actors and Positions

The California State Government is a leading group in regulatory tightening. The Newsom administration is attempting to internalize the external costs of data center expansion to operators through disclosure mandates and cost-sharing [18]. This is also a political choice by the state government to respond directly to voters' grievances over electricity rates.

Competing Host States (Georgia, Texas, Ohio, etc.) are taking the opposite path. They are actively absorbing data center investments while maintaining tax incentives [3]. For these states, the incentives of job creation and tax revenue expansion carry more weight than the regulatory burdens of California-style policies.

Nevada Utility Companies chose to terminate supply contracts to California and focus on constructing data centers within their own state [15]. This is a rational response by operators pursuing regulatory arbitrage, as well as an empirical example of the balloon effect, where regulatory tightening shifts burdens to adjacent jurisdictions.

Frontier AI Companies are pursuing multi-trillion-dollar data center investment plans, driven by the conviction that expanding computing power is the key path to improving AI performance [5]. For them, local backlash and state-by-state regulatory variations are risk variables that directly affect investment speed and site selection.

Local Residents and Bipartisan Public Opinion Coalitions are organizing opposition movements centered on electricity rates, water, noise, and safety accidents [5][17]. For them, the climate crisis is a secondary frame, and everyday grievances are the primary resource for mobilization.

Consulting and Policy Research Institutions (such as Bain, Roland Berger, and the Carnegie Endowment for International Peace) point out the perception gap surrounding the actual scale of AI's energy consumption. Bain revealed that while executives and consumers overestimate AI's share of electricity consumption at 11% and 19% respectively, its own modeling estimates it at a much lower level of around 0.7% [8]. This perception gap is a factor that destabilizes the very foundation of policy debates.

4. Key Issues

The first issue is the framing of the backlash itself. While public opposition to data centers is spreading, local media outlets, including *Wired*, commonly point out that this is not translating into discussions on addressing the climate crisis [9]. As long as immediate grievances over electricity rates, water, and safety take precedence, structural agendas such as emissions reduction or transitioning to a decarbonized grid are highly likely to be pushed down the political agenda.

The second issue is the balloon effect created by state-by-state regulatory variations. As California-style regulations spread, operators have incentives to relocate their investments to competing host states such as Georgia, Texas, and Nevada [3][15]. This has the potential to result in the environmental and energy burdens of data centers being transferred between regions rather than being reduced in total within the United States.

The third issue is the implications of this phase for South Korean companies' investment strategies in the United States. Given the differing regulatory intensities and incentive structures across states, South Korean companies need to establish state-level response matrices and constantly monitor regulatory variations, rather than relying on a single nationwide strategy [3]. Furthermore, it is crucial to separately track signals of a phase shift where the backlash merges with climate discourse—specifically, when the frame centered on electricity rates and water expands into emissions regulations [3].

II. In-Depth Analysis

The Clash Between AI Data Center Expansion and Climate Issues: In-Depth Analysis

1. Root Cause Analysis

The root cause of this issue is the time lag between the scale of AI infrastructure investment and the speed of power grid expansion. Frontier AI labs are convinced that expanding computing power is the key path to performance improvement, and are pouring trillions of dollars into data center construction [5]. Canada's TechInsights defined this as a phenomenon where 'data center emissions outpace the speed of grid decarbonization' [2]. This means that the physical expansion of the digital economy is faster than the pace of the clean energy transition.

Rising electricity rates are the most direct channel through which local residents feel this time lag. The Center on Global Energy Policy at Columbia University's SIPA analyzed the drivers of rising electricity rates across the US, identifying the surge in power demand from AI and data centers as a key variable [13]. EAI's existing analysis concluded that this backlash 'is being organized on a bipartisan basis around rising electricity rates, water usage, and safety accidents, rather than the climate agenda' [3]. In other words, the root cause is not a gap in perception regarding the climate crisis, but rather a distributional issue of who bears the costs of infrastructure expansion.

There are also structural reasons why this distributional issue is separated from climate discourse. According to a survey by Bain & Company, corporate executives estimate that AI will consume about 11% of global energy within three years, while general consumers perceive this figure to be even higher, at 19% [8]. In contrast, Bain's own climate-economics model predicts the actual share to be 0.7% [8]. This indicates a significant gap between perception and actual measurements. As long as this gap exists, the focus of the backlash is bound to be aligned with immediate and visible daily inconveniences—such as electricity bills, tap water supply, and noise—rather than macro indicators like greenhouse gas emissions. *Wired* pointed out this phenomenon, noting that 'backlash against data centers should also lead to reflection on the climate crisis, but it has not yet done so' [9].

2. Structural Context

Political Structure: Local Alliances Beyond Ideology

The backlash against data centers cannot be explained along the traditional left-right ideological axis. The Canadian outlet *The Globe and Mail* commented that 'the left and the right have united in their fear and loathing of data centers' [17]. The Brookings Institution stated that the data center debate has emerged as a national concern in both small towns and large cities, driven by opaque land-use agreements, rising electricity rates, environmental issues, and employment concerns [6]. The Carnegie Endowment for International Peace pointed out that lawmakers are emerging at the national level to demand a moratorium on data center construction [10]. This means that a place-based alliance transcending political factions has formed, which is a factor that increases political impact ahead of the 2026 midterm elections.

Economic Structure: Interstate Regulatory Arbitrage and the Balloon Effect

Under the US federal system, data center regulations are determined at the state level. This structure creates regulatory arbitrage. California, through seven bills signed by Governor Newsom on September 21, mandated water and power usage disclosures and required operators to bear the costs of grid upgrades [18]. In contrast, Georgia, Texas, and Ohio are absorbing investments while maintaining tax incentives [3]. This divergence has already led to actual shifts by operators. A prime example is a Nevada-based utility company terminating its contract to supply power to residents of South Lake Tahoe, California, to focus on constructing data centers within Nevada [15]. This indicates that the balloon effect, where investment shifts from highly regulated states to those with weaker regulations, is already in operation.

Security and Environmental Structure: Fossil Fuel Dependence for Emergency Power

The fact that large diesel generators and storage tanks have become standard equipment to guarantee the 'uninterrupted operation' of data centers is also a structural factor. Reporting on the diesel spill at the Equinix NY2 data center in Secaucus, New Jersey, *Caixin* commented that the facility, once a symbol of high technology, is degenerating into a potential source of petrochemical pollution [16]. In response to these concerns, the data center industry is pledging to reduce water usage, mitigate noise, and prevent cost-shifting, but experts tracking the industry worry that some solutions merely substitute one problem for another [14].

3. Historical Precedents and Comparison of Similar Cases

The structure of this backlash is similar to past conflicts over the siting of power plants and transmission towers. DW placed data centers alongside coal-fired power plants and wind turbines, pointing out that it is a classic NIMBY (Not In My Back Yard) structure where people 'do not want them in their own backyards but still want to use the services' [1]. This is a classic conflict over the siting of public goods, where benefits are dispersed nationwide or globally while costs are concentrated in specific localities.

The European cases serve as precedents where this conflict yielded political consequences at the local level before it did in the United States. In the French village of Étrelles, opposition to Google's plans to build a massive data center escalated to the point where an independent mayor resigned, and opposition movements were also organized in Kronstorf, Austria [7]. However, while European resistance was largely confined to sporadic local movements and data sovereignty discourse, the United States represents a more advanced stage, as seen in California, where the backlash is being institutionalized through state-level legislation [18][7].

Expanding further geographically, controversies have also emerged in Costa Rica over the entities constructing AI data centers in the Limón and Bagaces regions, suggesting that this conflict structure is spreading to developing countries as well [19].

4. Key Variables Shaping Future Developments

The first variable is whether state-by-state regulatory variations will persist. EAI analysis projects that the coexistence of California-style regulatory tightening and Georgia/Texas/Ohio-style investment attraction 'will not fundamentally change over the next one to two years' [3]. As long as this parallel landscape is maintained, the shift in operators' jurisdictions will continue.

The second variable is the signal of a phase shift where the backlash merges with climate discourse. Currently, the backlash remains confined to everyday issues such as electricity rates, water, and safety accidents [3]. The fact that discussions at Climate Week in New York addressed how AI's energy consumption will be integrated into the climate action agenda [12] shows that this integration is being attempted first on the international climate diplomacy stage. However, for this integration to materialize in US domestic politics, a catalyst is needed that clearly attributes the responsibility for rising electricity rates to the AI industry.

The third variable is the extent to which the issue becomes politicized during the midterm election cycle. The key is whether the demands for a nationwide construction moratorium, as pointed out by the Carnegie Endowment for International Peace, will actually lead to legislative proposals [10]. This serves as a test of whether place-based bipartisan alliances can translate into federal-level policy variables within the US domestic political landscape.

The fourth variable is the linkage with the US-China AI competition. The US race to secure data centers is intertwined with its strategy to maintain AI hegemony through infrastructural dominance [11]. If permitting delays or rising costs caused by local backlash slow down the US in this competition, the federal government may have greater incentives to intervene in state-level regulations. In this scenario, the possibility cannot be ruled out that political priorities will be reorganized in a direction where climate and environmental regulations are subordinated to the logic of national security and technological hegemony.