Hot Data: What the AI Boom Is Doing to the Land Around Us
Updated
Nobody asked the people living within six miles of an AI data center whether they wanted to be warmer. Nobody asked the deer, the pollinators, the soil, or the migratory birds either. Nope. They didn’t get a vote. A new study has put numbers to something that was hiding in plain sight. Researchers analyzed more than 6,000 AI “hyperscale” data centers worldwide. What they found is frightening. They discovered that after a facility begins operations, land surface temperatures in the surrounding area rise by an average of 3.6 degrees Fahrenheit. In the worst cases, nearby temperatures climb by up to 16.4 degrees Fahrenheit. The warming doesn’t stop at the fence line either — it radiates outward for up to 6.2 miles, affecting an estimated 340 million people globally. The researchers call it the “data heat island effect,” and they say it is real, consistent, and happening everywhere these facilities exist.
Again, this is not a theory. Researchers of the study — from the University of Cambridge and led by Andrea Marinoni of the Earth Observation group — deliberately filtered out seasonal effects, global warming trends, and other external variables, drawing on 20 years of NASA satellite temperature data. They focused only on data centers sitting away from dense urban areas, so nothing else could be blamed. What was left was the heat signature of the buildings themselves — and it is impossible to ignore.
In a world being quietly cooked, the mechanics are straightforward enough. A single hyperscale data center — the kind required to run AI models — can consume anywhere from 100 to 300 megawatts of electricity. That energy (like all energy) does not vanish. Instead, it becomes heat. The servers generate it, the cooling systems push it outside, and the backup generators add more. All of it lands in the surrounding environment, day after day, year after year.
In Mexico’s Bajio region, which has quietly become a data center hub, researchers documented unexplained temperature increases of around 3.6 degrees over the last two decades. Aragon, Spain — a major European hub for AI hyperscalers — recorded the same increase, with no similar rise in neighboring provinces. The pattern is the same wherever these facilities cluster: the land gets hotter, and the things living on it have no choice but to adapt or leave.
Temperature increases of that scale alter soil chemistry. They disrupt plant growth cycles. They push wildlife out of habitats that have sustained them for generations. On top of the heat, data centers bring noise — the constant hum of server fans and HVAC systems — which disrupts animal communication, alters behavior, and forces species to abandon territories. Noise, heat, and habitat destruction from facilities that can sprawl across 500 to 1,000 acres of what was previously forest, farmland, or open space. The land does not come back from that quickly.
Neil Carter, an associate professor at the University of Michigan’s School for Environment and Sustainability, has described data centers as potential “sensory danger zones” for wildlife — places where light and noise exceed the thresholds animals need to hunt, communicate, and reproduce. His research has shown that light pollution alone can throw off the hunting and survival dynamics between predators and prey — cougars and mule deer being one example. Add relentless heat to that equation, and what you get is not a thriving ecosystem. You get animals that cannot hunt, cannot communicate, and have nowhere left to go that isn’t compromised.
For anyone who has spent the last decade being told that individual carbon footprints and gas stoves are cooking the planet, here is something worth sitting with. Here is the wrinkle. Data centers are already keeping fossil fuel power plants running past their projected lifespans simply to meet energy demand. Capital expenditures for these facilities are predicted to hit $760 billion in 2026 alone, up from $450 billion the year before, according to BloombergNEF. Google’s parent company, Alphabet, plans to pour $185 billion into AI infrastructure — a figure that exceeds the GDP of entire countries. Corporate emissions targets have suffered repeated setbacks specifically because of data center expansion.
And now there is this: the facilities themselves, separate from their carbon output, are physically warming the land around them. The Cambridge study notes that these findings are particularly alarming because data centers are set to multiply rapidly over the next several years — even as heat waves are already becoming more extreme. The researchers were careful to filter out existing climate trends before measuring data center heat — meaning what they found is in addition to whatever else is already warming the planet. One warms from the top down—the other bakes from the ground up.
Deborah Andrews, emeritus professor of design for sustainability at London South Bank University, put it bluntly after reviewing the study: the rush for AI has been “developing far more rapidly than any broader, more sustainable systems.” That is a polite way of saying that nobody with the power to slow this down is slowing it down.
It is worth noting who tends to absorb these costs. Data centers disproportionately land in lower-income communities and communities of color — areas that are already carrying a heavier pollution burden and have the least political leverage to push back. The jobs promised during construction largely disappear once the buildings are running. A U.S. Chamber of Commerce report found that while a data center employs an average of 1,688 workers during construction, it typically provides just 157 permanent positions once operational. The heat stays. The jobs do not.
Communities are starting to fight back, though. In New Brunswick, New Jersey, residents successfully forced the cancellation of a proposed data center project after raising environmental concerns. New York lawmakers are considering a three-year statewide moratorium on large new data center projects while environmental and energy impacts are studied. In Virginia, the Data Center Reform Coalition has organized around five pillars of reform, pushing for mandatory disclosure of energy use, water consumption, and emissions, along with state-level oversight and ratepayer protections so that residents are not quietly subsidizing billions in infrastructure costs for the industry.
These are not fringe movements. They are people who live near these things and can feel what is happening.
So, where does this lead if nobody pumps the brakes? The Cambridge researchers themselves acknowledged there may still be time to consider a different path — one that doesn’t sacrifice the benefits of AI but demands that the infrastructure supporting it be held to an environmental standard. Marinoni said as much directly. The problem is simple: the money is too big. Alphabet, Microsoft, and the rest are spending at a scale that makes regulation feel almost quaint. Nobody in a position to slow this down has shown much interest in doing so. There are no enforceable thermal output standards. There is no serious national framework for where these things get built or what they are allowed to do to the land around them. So, they keep going up, and the heat keeps spreading.
The people living six miles from one of these facilities did not sign up for this. Neither did the animals. Nobody knocked on doors and explained that the neighborhood was about to get measurably hotter so that an AI chatbot could answer questions faster. The 340 million people sitting inside that warming radius are finding out the way people always find out — after the fact, when the damage is already done. At some point, that has to be part of the extremely important conversation. Sadly, it isn’t yet.
A clear, well-articulated argument against a new data center in Ohio… worth watching pic.twitter.com/c4hfr3EaBf
— non aesthetic things (@PicturesFoIder) April 12, 2026