Will the data-centre backlash derail the AI boom?

The rapid expansion of generative artificial intelligence faces a physical bottleneck as local planning boards and utility providers struggle to accommodate the massive energy and water requirements of hyperscale data centers. While AI software evolves weekly, the physical infrastructure required to run it takes years to permit and build, creating a growing friction between tech ambitions and municipal limits.

Data centers are the physical backbone of the AI boom, but their appetite for electricity and water is triggering local opposition and straining aging power grids. This “physical layer” of AI is now the primary risk to the industry’s growth trajectory, as the speed of software innovation outpaces the speed of electrical grid upgrades and zoning approvals.

According to the International Energy Agency (IEA), electricity consumption from data centers, AI, and the cryptocurrency sector could double by 2026. The IEA estimates that data center electricity consumption grew from 460 terawatt-hours (TWh) in 2022 to an estimated 800 TWh in 2024. This surge is driven by the transition from traditional search queries to generative AI, which requires significantly more compute power per request.

Tech giants including Microsoft, Google, and Amazon are now facing a “NIMBY” (not in my backyard) backlash. Local governments in the United States and Europe are increasingly rejecting or delaying permits for new facilities, citing concerns over noise pollution, land use, and the diversion of critical resources from residential areas. This clash represents a fundamental conflict between the Silicon Valley ethos of “moving fast and breaking things” and the slow, deliberative nature of municipal planning boards.

Why power grid capacity is limiting AI growth

The primary constraint on AI expansion is not the availability of chips, but the availability of electricity. Modern AI models require specialized GPUs that consume far more power than traditional CPUs. This has led to a surge in demand for “hyperscale” data centers—massive facilities that can house tens of thousands of servers.

In many regions, the electrical grid is unable to handle these sudden spikes in demand. In Northern Virginia, the world’s largest data center hub, Dominion Energy has warned that the pace of growth is testing the limits of the transmission system. The company has had to implement stricter requirements for new customers to ensure grid stability, according to Dominion Energy’s regulatory filings. When a single data center campus requires hundreds of megawatts of power—equivalent to the needs of a small city—the local grid often requires multi-year upgrades to substations and transmission lines before a facility can go online.

This “interconnection queue” has become a major hurdle. In the U.S., thousands of energy projects, including renewables intended to power data centers, are stuck waiting for approval from regional grid operators. The Federal Energy Regulatory Commission (FERC) has acknowledged these delays, which can stretch from three to five years, effectively capping the speed at which AI infrastructure can be deployed.

How water scarcity and environmental costs trigger local backlash

Beyond electricity, the cooling requirements of AI servers are creating environmental conflicts. Data centers generate immense heat and typically rely on evaporative cooling systems that consume millions of gallons of water daily. This has led to direct conflicts with agricultural interests and residential water security, particularly in drought-prone regions like Arizona and Texas.

Microsoft’s 2024 Environmental Sustainability Report revealed a significant increase in water consumption, partly attributed to the training and operation of large language models. The report showed a 34% increase in water consumption from 2022 to 2023, totaling billions of gallons. This transparency has provided ammunition for local activists and planning boards to challenge new permits.

In some jurisdictions, the backlash has moved from protests to policy. Local governments are increasingly requiring “water-neutral” plans or demanding that tech companies invest in wastewater recycling plants as a condition for zoning approval. These mandates increase the capital expenditure for data center operators and lengthen the time from site acquisition to operational status.

Can nuclear energy bypass the gridlock?

To avoid reliance on strained public grids and meet carbon-neutral goals, Big Tech is pivoting toward dedicated, “behind-the-meter” power sources. The most significant trend is a return to nuclear energy, which provides the constant, high-capacity “baseload” power that AI requires—something intermittent wind and solar cannot provide alone.

In a landmark agreement announced in September 2024, Microsoft entered into a 20-year power purchase agreement with Constellation Energy to restart a reactor at the Three Mile Island nuclear plant. The facility, renamed the Crane Clean Energy Center, is intended to provide a dedicated stream of carbon-free electricity directly to Microsoft’s data centers. This move allows Microsoft to bypass the public interconnection queue and secure its own power supply.

Similarly, Amazon Web Services (AWS) purchased a data center campus in Pennsylvania that is directly connected to the Susquehanna Steam Electric Station, one of the largest nuclear plants in the U.S. By owning the facility and the connection to the plant, Amazon minimizes the risk of grid-related delays.

The industry is also betting on Small Modular Reactors (SMRs). These are smaller, factory-built nuclear reactors that can be deployed closer to the data centers themselves. While SMRs promise a scalable solution, they remain largely in the developmental and regulatory approval phase. The U.S. Nuclear Regulatory Commission (NRC) maintains a rigorous licensing process that means wide-scale SMR deployment is unlikely to alleviate the power crunch before the late 2020s.

The economic risk to AI valuations

The friction between AI’s digital needs and physical reality creates a potential valuation risk for the tech sector. Markets have priced in an exponential growth curve for AI services, but that growth assumes that the physical infrastructure can be built as quickly as the software is written.

If planning boards continue to block sites and utilities cannot upgrade grids fast enough, the “AI boom” may hit a plateau. This would create a supply-side constraint where the demand for AI compute exists, but the physical capacity to provide it does not. Analysts suggest this could lead to a “compute shortage,” driving up the cost of AI tokens and slowing the rollout of next-generation models that require even larger clusters of GPUs.

Furthermore, the cost of securing power is rising. The shift toward dedicated nuclear deals and the construction of private energy infrastructure significantly increase the cost of operating AI. This puts pressure on the “ROI” (return on investment) of generative AI, as companies must spend billions on power and cooling before they can monetize the software.

Global variations in data center regulation

The backlash is not uniform globally, and this is leading to a shift in where AI infrastructure is built. In the European Union, the “AI Act” and strict environmental directives are forcing a more cautious approach to data center expansion. Countries like Ireland and the Netherlands have previously implemented moratoriums or strict limits on new data centers to protect their national grids.

In contrast, some Southeast Asian nations and Middle Eastern countries are offering aggressive incentives, including subsidized land and energy, to attract hyperscale operators. This “regulatory arbitrage” may shift the center of gravity for AI compute away from traditional hubs in the U.S. and Europe toward regions with fewer zoning restrictions and more available land.

However, these regions face their own challenges. The extreme heat in the Middle East increases the energy required for cooling, potentially offsetting the benefits of cheaper land. The physical laws of thermodynamics remain the ultimate constraint, regardless of the political environment.

What happens next for the AI infrastructure race

The industry is moving toward a model of “energy-first” site selection. Rather than choosing a location based on taxes or talent, tech companies are now choosing sites based on proximity to high-voltage transmission lines or existing power plants. This is transforming the real estate market for industrial land, with “power-ready” sites commanding massive premiums.

The next critical checkpoint for the industry will be the upcoming quarterly earnings reports and capital expenditure guidance from Microsoft, Alphabet, and Amazon. Investors will be looking for evidence of how these companies are managing the “physical bottleneck”—specifically, whether they are increasing spending on energy infrastructure and how they are navigating the permitting delays in key markets.

Additionally, the U.S. Department of Energy is expected to provide updates on the deployment of SMRs and the modernization of the national grid. Any acceleration in federal permitting for transmission lines could alleviate some of the pressure on local planning boards, but the fundamental tension between industrial-scale compute and local community interests is likely to persist.

Share your thoughts on the AI energy crisis in the comments below or share this analysis with your network to join the conversation on the future of digital infrastructure.