In June 2024, Microsoft published its ‘Datacenter Community Pledge’, addressed to the communities its data centres affect. It states: “We owe a commitment to be responsible neighbours and contribute positively to local economies and ecosystems while advancing digital transformation.” The pledge continues, “We will design and operate our datacenters to support society’s climate goals and become carbon negative, water positive and zero waste before 2030.” It also promises to procure 100 percent renewable energy by 2025.
So why, in public Microsoft reports from the same year as the pledge was published, was renewable energy consumption omitted? Previous reports cited that this was confidential information. Water usage and total energy consumption were not reported at all. If community is such a priority, why has Microsoft failed to consult local residents about plans for its site in Person County, North Carolina, in the 18 months since they bought the land?
Microsoft is competing against its Big Tech counterparts—Google, Meta, Amazon, Musk’s xAI—to dominate the AI landscape. The corresponding boom in data centres from a world desperate to keep up with AI trends leaves citizens and the environment suffering the consequences. A shift in the narrative is desperately needed, away from devastating impacts on communities and the climate towards a sustainable digital future.
Speed of innovation trumps public health
In Memphis, Tennessee, Elon Musk’s data centre (run by his company, xAI) is powered by 35 methane gas turbines. The turbines emit nitrogen oxides, producing smog, which is then inhaled by local citizens. “Right now, what xAI is doing is essentially running a power plant without a permit,” according to Amanda Garcia, senior attorney for the Southern Environmental Law Centre. XAI has no Clean Air Act permits or pollution controls (it is exempt from the usually mandatory requirements because its turbines are only ‘temporary’). “It has a real risk of harming people’s health,” Garcia added.
More chilling still is exactly whose health the smog is damaging. Boxtown, a neighbourhood with a 90 percent Black population, is just three miles from xAI’s data centre. Historical racism remains prevalent in Memphis’s 1953 zoning maps, which are still used to plan the city. Just as AI models can be racist due to racial bias in algorithms, the negative health impacts of AI models are disproportionately felt by the Black community. Similarly, Indigenous land is consistently used for data centre builds, as monitored by Honor Earth’s data centre tracker. Data centres’ degradation of ecosystems is particularly damaging for Indigenous communities who rely upon these ecosystems for survival.
A concerning chemical comes to light
PFAS, a chemical which has been linked to cancer, birth defects and other serious health problems, is used by data centres in their cooling systems, pipes, cables and semiconductors. Climate advocates are calling out the unreported use of PFAS by data centres, while the industry dismisses concerns, insisting that pollution from PFAS isn’t a threat to human health or the environment.
XAI is using turbines because local utilities can’t fulfil its energy demands (which are equivalent to those of a city with 80,000 inhabitants). Musk deemed the turbines necessary for xAI to accelerate the launch of the third version of its chatbot. In the race for AI dominance, speed is a crucial factor. In Musk’s eyes, public health is secondary to beating his competitors.
GenAI demand is driving data centre build-out
It’s not just Big Tech titans locked in the battle for AI dominance. GenAI use is soaring worldwide, and the number of AI-ready data centres is booming as companies race to keep up with growing demand. Analysis from McKinsey predicted that by 2030, around 40 percent of total data centre demand will be for data centres equipped to handle GenAI workloads.
Training GenAI models requires GPUs, which are able to process several data streams simultaneously. This is so energy-intensive that some GPUs, such as the chip made by market-dominating Nvidia, get so hot that they can’t be cooled with air conditioning alone. As GenAI models become increasingly commonplace in everyday life, the energy use of data centres—and the energy demand of entire cities—will skyrocket in response.
Energy grids are under pressure while consumer bills rise
Thanks to this surge in energy demand, several data-centre hubs are experiencing increased pressure on their local energy grid. In 2023, Ireland’s data centres consumed more electricity than its urban homes. Patrick Brodie, Assistant Professor at University College Dublin, told RESET that, “There is well-founded speculation that [Ireland’s data centre energy demand] is leading to higher energy bills for consumers.” Meanwhile, in the US, analysis by the Lawrence Berkeley National Laboratory predicts that data centres will account for between seven to 12 percent of the nation’s electricity demand by 2028. As reported by Bloomberg, there has been a 267 percent rise in wholesale electricity costs near US data centre hubs. And worldwide, the International Energy Agency foresees data centres’ global electricity consumption matching that of Japan by 2030.
Besides straining the grid and increasing energy bills, data centre growth is also jeopardising the energy transition. Firstly, by increasing our power demand, and secondly, because we’re unable to reduce our reliance on gas, as seen by an increase in proposed gas plants across Europe. This extreme growth in energy consumption “makes it more difficult to decarbonise,” according to Brodie. He tells us how the data centre industry in Ireland is using natural gas as a transition fuel to accommodate their expansion, which in turn increases their emissions. Research from the International Energy Agency shows that fossil fuels currently meet almost 60 percent of data centres’ energy demand. And from the UK to Germany, there are reported cases of data centres that planned to use renewable energy, before turning to fossil fuels to fill the gap.
Data centre cooling systems are draining vital water resources
With energy a huge problem in data centres, it’s unsurprising that Big Tech companies will shout loudly about any reduction in their energy use. But a common reason for reduced energy in data centres is replacing air conditioning with water-based cooling systems. While rainwater might seem the obvious choice for cooling, many data centres use drinking water to reduce costs associated with sterilising rainwater. One data centre requires the equivalent daily water supply as a town of 10,000 to 50,000 people. AI-heavy workloads are especially thirsty; training ChatGPT-3 consumed as much as 700,000 litres of freshwater, according to a 2023 study. In drought-stricken countries, data centres are draining water resources needed by humans in an increasingly dystopian battle between Big Tech companies and local citizens. “Both [water and energy] are ecological problems,” says Brodie, “as one single industry is increasingly monopolising demand and amplifying their usage unsustainably to maintain profits.”
Recent developments and projections for yearly global AI energy consumption A shift is needed away from the AI hype
The majority of Big Tech companies—and their associated data centres—might be based in the US, but it’s clear Europe doesn’t want to be left behind. In early 2025, the EU launched a program with a €200 billion budget. Dubbed ‘InvestAI’, it plans to build five AI gigafactories “to make Europe an AI continent” and “develop and train complex AI models at an unprecedented scale”.
The need for more AI models may not be motivating this investment as much as the fear of being left behind. RESET spoke to Jill McArdle and Pierre Terras from Beyond Fossil Fuels about the current AI hype. They told us that, “[Big Tech] are actively promoting the integration of generative AI across all sectors of life—public and private, personal and professional—with little regard for whether we can sustain the infrastructure behind this, or whether this type of AI is truly needed across all these areas.” As stated by Sasha Luccioni, AI and Climate Lead at Hugging Face, in an interview with the Financial Times, “It’s almost like a mass hallucination where everyone is on the same wavelength that we need more data centres without actually questioning why.”
McArdle and Terras also pointed out that, “We are seeing a growing gulf between Big Tech companies’ climate commitments and their AI expansion strategies.” Google and Meta claim they will reach net zero by 2030, while Amazon has given itself until 2040 to achieve the same goal. Microsoft, meanwhile, plans on being carbon negative by 2030. Whether these goals will be achieved, superficially met through carbon credits, or disregarded entirely remains to be seen. But, with devastating consequences for people and the environment, building more and more data centres is at odds with the very idea of pledging to help the climate.
Do data centres boost local economies?
Another mass hallucination at play is how Big Tech claims that its data centres benefit communities by boosting local economies. But a study commissioned by Germany’s federal economic ministry showed that the majority of investment was spent buying hardware and constructing buildings. Data centres “most often create value far from where they are built,” according to Max Schulze, who carried out parts of the study.
Tougher legislation has the potential to spark change
One solution for minimising the environmental impact of data centres is tougher legislation. In Germany, the Energy Efficiency Act, which came into effect in November 2023, requires that data centres be powered by 100 percent renewable energy by 2027. It also stipulates that data centres meet Power Usage Effectiveness and Energy Reuse Factor figures; they can hit these targets by reusing waste heat from their operations. However, reusing waste heat isn’t as simple as connecting a few pipes. According to Dr Ralph Hintemann from the Borderstep Institute, it’s only a worthwhile endeavour when building new data centres, as “replacing an existing heat supply is [too complex].”
Meanwhile, in Italy, the Ministry of the Environment encourages tech companies to build data centres on brownfield sites (areas already used for development) rather than in green spaces. But encouragement alone isn’t enough to lead to change; the guidelines are often ignored, with no ensuing penalties.
So, what regulation is needed? McArdle and Terras tell us, “It’s the responsibility of policymakers and regulators to set sustainable limits on AI’s infrastructure’s consumption of resources like energy and water.” They must also ensure that “data centres are not given priority access to the grid without any social or environmental criteria.” Beyond restrictions on data centres, “Policy makers must begin to assess the proven societal benefit of different AI applications and to deprioritise non-essential uses of this resource-intensive tech.”
According to a 2025 survey of over 5000 participants across five European countries, 64 percent of respondents believe data centres shouldn’t be built if they will be powered by fossil fuels, and 85 percent demanded they disclose their environmental impact. This indicates a growing public awareness of data centres’ damage: exactly the shift that’s needed to put pressure on policymakers and Big Tech platforms alike.
Make your voice heard with a data centre protest
Citizens have the power to stop data centres in their tracks. Read Algorithm Watch’s guide for advice on resistance for local communities.
Can data centres exist without harming people and the planet?
Beyond Fossil Fuels’ data centre report states that, “The growth of data centres risks derailing Europe’s energy transition, unless immediate action is taken.” We posed the question to McArdle and Terras: What action must we take?
Their demands firstly fall on data centres, which they stipulate must be fossil-free and use renewable energy. Here, they specify that companies must “move away from offsetting their energy use with certificates to being 100 percent matched with renewable energy around the clock.” They also demand sustainable limits on growth and a true commitment to transparency from data centre companies. (Transparency in the data centre industry is often a fallacy. Microsoft’s 2024 reports, for example, omitted renewable energy consumption on the grounds it was confidential, while the water input and total energy consumption fields were left blank.)
So, what could a truly sustainable data centre look like in practical terms? A great example comes from Swiss company Infomaniak. Its latest data centre is located beneath a public park in Geneva, where it redirects its waste heat to local buildings. Its cooling system is completely circular (they redirect cold air from their heat pumps to cool their servers), omitting the need for water-intensive cooling. It also upgrades broken or outdated servers to counter e-waste and runs on renewable energy.
Meanwhile, in Luxembourg, supercomputer MeluXina also runs completely on renewable energy and diverts its waste heat. It has opted for a closed circuit cooling system that has been reusing the same water for several years, powered by a biomass power plant fuelled by waste wood.
How can we ensure a green digital future?
Growing e-waste, carbon emissions from AI, data centre water usage—is rampant digitalisation compatible with a healthy planet? Our latest project explores how digital tools and services can be developed with sustainability in mind.
McArdle and Terras point out that “data centres can be very small. A 2MW data centre hosting data for a local municipality is not an issue.” Similarly, small, community-owned data centres give citizens sovereignty over their data and empower them to decide for themselves how a data centre will co-exist with their local land. For users, relying less on Big Tech products is another way to mitigate their harm, while developers must also follow sustainable AI development guidelines. “There may be smaller task-specific AI models that could perform many of the functions of LLMs with far fewer resource inputs,” say McArdle and Terras. “Some data centres are necessary,” they continue, “but we also need to question the necessity of the explosion in AI infrastructure we are seeing today.”
