Content
Table of Contents

Why We Need a Moratorium on Hyperscale AI Data Centers

AI tech giants are establishing “data farms” across America and the demand for hyperscale data centers is expected to triple by 2030. Much of this new construction is being carried out on rural and agricultural lands. This has led to concerns i regards to the infrastructure and resources required to support hyperscale data centers, and the impacts on human health, water quality, air quality, biodiversity, and cultural identity.

Hyperscale data centers are not your traditional data centers of the 2000s - 2010s that made social media possible or served as data storage for academic or health institutions. These are factories - with at least 10,000 sq ft and more than 5,000 servers, being used for e-commerce, cloud computing, and, more often, AI (including your Language Learning Models [LLMs]. and data surveillance). This complex workload comes with real risks to our communities.

Energy Risks

Hyperscale data centers consume a massive amount of energy because of those complex workloads. A single 100MW data center can consume as much energy as about 100,000 US households.

In 2023, data centers consumed about 4.4% of the total electricity used in the US and by 2028 that is expected to grow to between 6.7-12%. Most of that energy is provided by local power grids, but there is a new trend for developers to include independent or on-site power sources, including gas-fired power plants.

Additionally, the primary source of the energy is fossil fuels. In 2025, it is estimated that data centers in the US emitted more than 52 million tons of carbon dioxide into the atmosphere. Based on current trends, it’s projected that new data center developments could emit 12.1 billion tons of carbon dioxide. This is a huge concern in the context of climate change when we consider the fact that we are rapidly approaching what the IPCC has long recognized as the tipping point. We are also feeling the effects of climate change right here in Polk county, including higher temperatures, greater humidity, more intense rainfalls (which ironically don't contribute much to groundwater recharge), and more mosquito-borne illnesses (hello Dengue!).

Water Risks

All of those servers generate a lot of heat, and so about 40% of a hyperscale data center’s energy demand is used for cooling. According to the EPA, a single hyperscale data center can use 1-5 million gallons of water/day for cooling - equivalent to roughly 16,000 US households. By 2030, it’s estimated that hyperscale data centers will consume 33 billion gallons of water annually.

Now, about 75% of hyperscale data centers are AI data centers, powering the LLMs. To put the impacts of using ChatGPT or Claude into perspective, a single 100-word prompt uses, on average, 519 milliliters of water, which is the equivalent of a standard plastic water bottle. This high water demand by data centers, understandably, reduces water available to communities as a whole and impacts water flow and pressure to nearby homes. A reduction in water pressure and sediment in the water is a common report from families that live near these facilities.

Concerns about water quality and contamination are also on the rise. Data centers release a huge amount of waste water and that either enters into a publicly owned treatment works (POTWs) or into direct water sources through National Pollutant Discharge Elimination System (NPDES) permits. As data center waste water discharge increases, the current facilities will not be able to handle the high volume and will therefore, not be able to fully and effectively treat waste water, posing a possible public health risk.

The direct discharge, however, is a concern for multiple reasons. First, is that if a community already has poor quality water, the evaporative cooling systems will concentrate those contaminants and then the discharge water will put those concentrated pollutants directly into groundwater sources or surface waters. Second, is that data centers frequently use chemical additives, including heavy metal and biocides, to prevent scale or biofilms from building up in the lines. PFAs (aka "forever chemicals") is another chemical additive of concern with data centers. The CDC and EPA have both recognized these types of contaminants as health hazards - being known carcinogens and linked with reproductive effects, developmental delays, a variety of cancers, and hormone imbalances. Now for closed loop cooling systems, while water consumption and waste water discharge is less of a worry, they still pose a contamination risk whether through malfunction or routine maintenance, like flush-and-fill or bleeding-the-lines.

Air Quality

Within those data center emissions that I mentioned earlier is another concern: fine particulate matter or FPM. FPM is a known severe health hazard that is linked to respiratory diseases, particularly asthma.

Studies have found that residential communities living within one mile of a data center have comparatively worse air quality. Now, Polk County already does not have a good track record on this. In fact, Lakeland, specifically, was ranked as the 7th worst place to live with asthma in 2024 by the Asthma and Allergy Foundation due to poor air quality.

Researchers at the University of California have studied the direct relationship between air quality and AI data centers and found that emissions from these facilities could lead to 600,000 symptomatic asthma cases and 1,300 premature death. What’s more, another study used EPA modeling and their air operating permit data found that emissions from data centers could pose an estimated $53-99 million per year in health related damages. That is a significant public health burden.

Noise Pollution

Hyperscale data centers generate a lot of noise, owing to their HVAC systems and cooling technologies as well as their backup diesel generators that run 24/7/365. These standard facility operations can easily emit 70 dB within a 400 foot radius, but the noise has been found to emanate upwards of about 3000 ft. If it is a data center that produces its own power, and so has substations on site, that can actually have noise levels reaching 100 dB.

The Center for Hearing and Communication reports that chronic exposure to 70 dB as being linked to hearing loss and anything over 91 dB is considered dangerous. But long-term exposure to noise pollution is also associated with anxiety and depression, headaches, sleep disruption, and learning disorder in children. A new study from Brown University found a statistically significant decrease in the math scores of 3rd graders attending public schools within one mile of a data center.

Environmental Justice

Hyperscale data center have a siting pattern of being constructed in vulnerable communities - specifically low income, communities of color. More recently, there has been a shift to agricultural or rural areas. In both locations, hyperscale data centers pose an environmental justice issue. These area experience disparities in wealth, critical services, and health outcomes. For low income, communities of color, this is reflective of a legacy of redlining in which industrial activities and been brought into these communities and they experience disproportionate environmental and health burdens, including worse outcomes for climate change and higher rates of cancer. In rural areas, there are issues with poor quality water, which then increases the risk of contamination with waste water discharge. There is also the added issue of data centers taking over rural and agricultural lands, which contributes to food insecurity and affordability, and also an area’s cultural heritage.

Wildlife & Livestock Impacts

Everything mentioned above impacts non-human animals, but noise pollution is a massive concern. Noise pollution alters basic biology and behavior and can lead to both increased and earlier mortality of many animals. Additionally, noise pollution negatively impacts communication between animals, disrupting reproduction, foraging, movements, and the ability to detect predators. A 2026 study found that in the presence of noise pollution, species abundance decreased 40% and mortality rose by nearly 140%. This is something that people living near data centers have reported - the wildlife seems to just disappear.

There are some online resources that discuss electromagnetic fields (EMFs) in relation to hyperscale data center and state that farmers or ranchers near these facilities are reporting lower reproduction or lifeless calves; however, those reports have not been confirmed. But knowing that noise pollution has a significant impact on wildlife, it is not a stretch to assume that noise pollution probably impacts livestock too. Data centers do pose a concern to livestock in terms of competition for resources - specifically space and water. Raising livestock requires a lot of land and those lands are being bought left and right by developers. And then the data centers are using water sources that livestock need to survive.

Economic Impacts

A 2026 study out of Georgia Tech looked at 2,500 data centers and found that, on average, a data center costs $1 billion to construct. When operational, they provide a small boost to the employment rate (3.5%) and median household income (2%). But to reach these numbers takes time - it is not an immediate return. And that authors note that is a small benefit for such a cost.

Another important finding was that not every community received these benefits. The above benefits were provided by data centers built in metropolitan communities that had an existing infrastructure that could accommodate such a demand. Non-metropolitan communities, such as where they are being proposed in Florida, do not get the same pay off. In communities like ours, data centers employ few permanent worker, typically only 25-100 and the high-paying positions are not brought to data centers in smaller communities. While there is a meager decrease in the unemployment rate, any benefit there is offset by tax breaks given to the data centers. Additionally, these communities need to invest in infrastructure expansions and that bill is footed by ratepayers.

And that brings up to ratepayer protections. While Florida has SB484 to protect ratepayers, there are loopholes. First, it allows for one year NDAs. This lack of transparency means that communities may not know what’s being developed until it’s essentially too late. Second, are how the utilities charge the data centers. The utility has to submit an estimate (tariff) of what they would charge the data center for their energy consumption. There is concern that in this process, they could inflate the rate to developers and then quietly shift that to the general rate. Finally, SB484 doesn’t prohibit speculative demand. Developers can submit multiple parallel requests to municipalities. If one decides to bite and invest in an infrastructure expansion, but the data center ends up going to a different municipality, ratepayers can be stuck with that bill.

Let's also highlight the value of nature in Florida. In 2017, the FDEP reported that outdoor recreation and nature-based tourism brought in $145 billion to Florida's economy and supported roughly 1.2 million jobs. Nature brings in far more revenue than hyperscale data centers ever could. And it's important to remember: If a data center does bring in profits, those will not return to Florida's economy. Instead those profits will go to the out-of-state companies.

Bottom Line

I’m first and foremost a scientist - I want scientific advancement and I want the US to be at the frontlines of that. But how hyperscale data centers are being built right now is damaging and that is clear. It destroys our habitats, depletes our natural resources, and makes our communities sick while a few millionaires or billionaires reap large profits.

What we need is for these developers to embrace innovation, themselves, so that our technological advances are also sustainable.

~ Ashley Bowers-Macrander

President, Polk County Democratic Environmental Caucus