Power, Pollution and Policy: The Environmental and Community Costs of AI Data Centre Expansion

In the modern world, a quiet revolution is reshaping how we live, work, and think, driven by artificial intelligence. Behind the scenes of every chat, image, and prediction lies a massive physical infrastructure: the data center. These are not just server rooms; they are sprawling industrial complexes that require enormous amounts of electricity to run their computers and vast quantities of water to keep them from overheating. For years, the industry has relied on a single number to prove these facilities are efficient, a metric called power usage effectiveness. This number compares the total energy a building uses to the energy the computers inside actually consume. A lower number suggests the building is wasting less energy on things like cooling and lighting. However, as artificial intelligence has exploded in popularity, the demands on these facilities have changed. The computers now run hotter and harder than ever before, and the simple efficiency score no longer tells the whole story of how much energy is being consumed or what it is costing the world.

This shift has created a tension between the rapid growth of technology and the communities that host these facilities. In the United States, the state of Virginia has become the epicenter of this expansion, specifically in a region known as "Data Center Alley." Here, the landscape has transformed from green pastures into a dense forest of industrial buildings, hosting a significant portion of the world's computing power. As these facilities multiply, they are straining local power grids, draining water supplies, and generating noise and pollution that affect nearby residents. The question facing scientists, policymakers, and neighbors is whether the current rules are strong enough to protect the people living next door to these digital giants, or if the benefits of artificial intelligence are being paid for with the health and resources of a few local communities.

A new systematic review by researchers Claire Kennedy and Mahmoud Al-Kilani from FHNW University in Switzerland takes a hard look at this situation. They focused their investigation on Virginia, examining the environmental and social costs of the data center boom while analyzing the laws and regulations designed to manage it. The researchers did not just look at the technology; they dug into the legal records, studying hundreds of bills introduced in the state legislature between 2025 and 2026. Their goal was to see if the government was keeping pace with the industry's growth and whether the laws being passed actually protected the people and the environment, or if they were merely creating the appearance of action.

The study begins by challenging the industry's favorite tool for measuring success: the power usage effectiveness score. The authors argue that this metric has become misleading. While a data center might report a very efficient score, this number only tells us how well the building manages its own internal energy, not how much total energy it is using. A facility could be incredibly efficient at cooling its servers while still consuming enough electricity to power a small city. Furthermore, the researchers found that the companies running these centers often report this data in ways that hide the true scale of their consumption. They might average the numbers across all their buildings or release them with a significant delay, making it nearly impossible for outsiders to understand the real impact of a single new facility. This lack of clear, immediate data creates a fog that allows companies to claim they are being green while their actual energy use continues to climb.

As these facilities demand more power than the local electrical grid can provide, companies are turning to alternative solutions that often come with their own heavy costs. The researchers found that many data centers are now relying on backup generators that burn natural gas or diesel fuel to ensure their computers never stop running. In some cases, they are building massive off-grid power plants that operate independently of the public utility. While this keeps the internet running, it shifts the burden onto the surrounding environment. The study highlights that these generators emit pollutants that can harm human health, contributing to respiratory problems and increasing cancer risks for nearby residents. Additionally, the demand for water to cool the servers is so high that it threatens local water supplies, potentially leaving less water for farms and households.

The core of the researchers' work involved a detailed analysis of the legislative response in Virginia. They examined 66 specific bills introduced in the state legislature that dealt with data centers, generators, and high-energy facilities. They categorized these proposals to see if they were trying to gather information, impose strict rules, or offer financial incentives. The results were stark. The vast majority of bills that sought to protect communities, limit pollution, or ensure that data centers paid their fair share for grid upgrades failed to pass. Out of the dozens of proposals, only a handful became law, and most of those were weak. They tended to be studies or reports that asked for more information rather than laws that forced companies to change their behavior. For example, one successful bill required a study on how to reuse the waste heat from data centers, but it did not compel the companies to actually share that heat or prove it was feasible.

The researchers noted a distinct pattern in the laws that did pass versus those that were rejected. The bills that successfully became law were often voluntary, asking companies to report their data or offering tax breaks if they promised to be more responsible. In contrast, bills that tried to set hard limits, such as banning certain types of polluting generators or requiring data centers to pay for the upgrades they caused to the power grid, were frequently blocked. In several instances, bills that had passed both houses of the state legislature were vetoed by the governor, who argued that they would create too much red tape and slow down economic growth. The study suggests that the state government is prioritizing the expansion of the artificial intelligence industry over the concerns of local residents, effectively leaving communities with little power to say no to new construction.

One of the most concerning findings of the paper is the concept of "greenwashing." The researchers argue that the industry is using vague promises and selective data to make it seem like they are solving environmental problems when they are not. By focusing on efficiency scores that ignore total energy use, or by promising to use renewable energy that is not yet available, companies can present a friendly image while continuing to expand their operations with minimal oversight. The study points out that without independent, standardized data and enforceable rules, these promises are difficult to verify. The result is a situation where the financial and environmental costs of artificial intelligence are being borne disproportionately by the communities that host the data centers, while the profits and benefits flow to the companies and their investors.

The authors conclude that the current approach is unsustainable. The rapid growth of artificial intelligence infrastructure is outpacing the ability of local governments to regulate it, and the existing metrics are too flawed to guide meaningful change. They argue that relying on voluntary reporting and tax incentives is not enough to protect the environment or public health. Instead, they call for a new system of rules that requires companies to provide clear, detailed, and up-to-date information about their energy and water use. They suggest that these rules must be consistent across different regions so that companies cannot simply move their operations to places with weaker laws. Without these changes, the paper warns, the true cost of the digital future will continue to be paid by the people living in the shadow of the data centers, with little recourse to demand a better outcome.