During a radiant sunset in Memphis in May, Sharon Wilson pointed a infrared camera at a large AI hub to reveal a environmental hazard her eyes could not. Free from pollution controls, the gas-fired turbines powering an enormous AI data farms were releasing invisible fumes into the Tennessee sky.
“I was completely stunned,” said Wilson, who believes the facility was emitting more planet-heating gas than a sizeable energy plant. “A truly shocking volume of emissions.”
Around the same time, the facility’s core product was running riot on news feeds, repeating false narratives and making inaccurate statements. Commentators have highlighted generated images of offensive characters as an illustration of what methane was being consumed to produce. “What is the actual benefit here?”
Wilson is not alone in voicing this concern. Researchers are watching the AI boom with concern as it pollutes the natural world with carbon and the online sphere with dangers ranging from medical myths to AI-generated abusive content.
Some experts fear datacentres may derail the shift to a low-carbon future, adding an major challenge to the enormous challenge of preventing the planet from heating beyond critical thresholds. Others are sanguine about the energy costs, arguing they seem minor not just to other sectors, but also to the technology’s potential to transform our world.
The Surging Energy Demand
While Hannah Daly ran models for an leading energy organization, she spent minimal effort worrying about the carbon cost of computers. Cars, cows and home heating were among the major issues consuming the carbon budget.
But in one European nation, the computational hunger has reached levels that are increasingly alarming. Datacentres use one-fifth of the country’s electricity and are expected to devour nearly one-third within a few years. The quick proliferation of these chip-filled warehouses has surpassed the electricity grid's capacity, leading to effective bans on new grid connections.
“The path of enormous, exponential growth is what is so alarming,” said Daly. “It's unclear whether this is a real outlier or a precursor of what’s to come. But it’s definitely a warning story.”
Globally, datacentres consume approximately one percent of the world’s electricity but may soon demand much more. Their percentage of electricity demand in major economies is estimated to increase significantly by 2035. Some demand is being met through long-term agreements to buy renewable power, while some tech firms have signed deals to use nuclear energy.
But for the foreseeable future, non-renewable sources are set to be the primary source. In some regions, datacentres are concentrated in areas reliant on polluting power generation. In others, political support has framed them as justification for continuing with dirty energy. “The assumption that the lower cost of renewables alone will drive decarbonisation – that is insufficient,” said Daly. “If there’s a huge source of energy demand that wants to grow, it will land on these stranded fossil fuel assets.”
Assessing the Environmental Footprint of a Interaction
Does using chatbots to draft emails harm the planet? Tech companies have resisted pressure to provide transparent metrics on their AI energy footprints, but common estimates hover around a small amount of energy for a simple text query, increasing significantly for more complex tasks. One CEO recently asserted a typical AI query uses as little energy as running a lightbulb for a couple of minutes.
These numbers are very small compared to activities such as air travel, eating meat, or driving a car. However, concerned voices worry about the massive volume of the technology – with hundreds of millions of weekly users – and the fervour with which companies have embedded it into every aspect of digital life. The rise of AI agents is expected to add to the ongoing power consumption.
“My concern is is that we’re implementing these systems in such a way that we are not fully aware of the energy use,” said a climate lead at an AI company. “The current approach is on the assumption that it’s not a problem – or that if it is a problem it will somehow be solved – instead of planning for it.”
The Promise for Emission Reductions
What if AI could balance its energy debts by reducing emissions in other parts of the economy? That is the thesis put forward in several reports, which argued existing AI applications could cut emissions by far more than datacentres produced. Researchers have modelled scenarios in which AI would help to manage renewable energy, identify plant-based meats, improve EV components and nudge people toward climate-friendly choices.
“AI can speed up the deployment of those clean technologies by basically advancing their development along the curve of innovation and adoption,” said a policy fellow involved in the research.
The potential climate benefits carry large uncertainties – greater efficiency can lead to more use, and unintended consequences may undercut the gains. But examples already exist. Some companies report AI has helped optimize thermal management in datacentres by a significant percentage. Energy firms say AI has optimised wind turbine maintenance and performance, increasing operational efficiency. Solar farm operators have reduced downtime by using AI to detect faults.
Because other sectors are so carbon-intensive, researchers say, AI would only need to cut their emissions by a small fraction to cover the carbon cost of its computing. “In the power sector we are already seeing the results,” said one researcher. “Whereas other sectors are not there yet.”
The Flip Side: Enabling Fossil Fuels
Sustainability proponents are not the only ones to have noticed AI’s powerful capabilities. The fossil fuel industry is also a major adopter. Some former tech employees have expressed concern over their companies' work for oil and gas clients, arguing it enables more emissions than the tech sector produces itself. Partnerships have been announced with major oil companies, with the potential to expand production and streamline exploration.
Industry estimates suggest AI could boost technically recoverable oil and gas reserves and cut project costs. Industry leaders have stated that AI is embedded “in everything” and has increased productivity, leading to more drilling activity. The oil and gas industry also argues AI can cut its carbon intensity, for instance by analysing data to spot methane leaks. But critics say there is a gap between data analysis and real-world changes.
“They are using this as an justification to delay action,” said one campaigner. “Observing methane from space is not preventing methane.”
Perhaps even more concerning than the expansion of fossil fuel supply is the effect on consumption. Generative AI has been shown to create very successful advertisements, and the ease of production slashes the cost of promoting buying. The marketing industry is preparing for AI agents that could make decisions on a customer’s behalf.
“The common discussion is really focused on this false comparison between the energy used to run the technology and the positive use cases,” said a campaigner. “But it is risky to overlook the negative use cases.”
Demands for Caution and Rules
Some voices are calling to hit pause, at least until better rules are in place. International bodies and coalitions of environmental groups have proposed a moratorium on new datacentre development, citing damaging environmental impacts. Some national regulators have imposed conditions, requiring that a large majority of a datacentre’s power must eventually come from new renewable projects.
Different voices have called to push the sector to do good. A few countries have started to reference AI in climate legislation, obliging governments to promote digitalisation that aids decarbonisation. There have also been suggestions to levy fees on computing to generate resources needed for climate action.
Some campaigners are pushing for upcoming regulations to classify certain high-emission applications of AI as concerning. They also want investors to consider enabled emissions when evaluating companies.
In response to concerns, some AI companies have stated they give considerable thought to the best use of computing power, support sustainability goals, and believe AI will be instrumental in tackling climate change. Other tech firms argue the energy transition is complex and requires a pragmatic strategy, with technology playing an important role in helping industries decarbonize.
One researcher concluded that, rather than panic, people should pressure companies to make tools that are energy-efficient from the start.
“I choose to be optimistic, but I still believe that AI can do good in terms of fighting the climate crisis – creating the next generation of batteries, tracking deforestation, predicting hurricanes,” she said. “Countless good things for which we can be using it – and instead we’re creating websites filled with frivolous outputs while datacentres are getting powered by diesel generators.”
