How big a threat is artificial intelligence to the climate

Sharon Wilson, a former oil industry worker and longtime methane leak observer, has used a thermal imaging camera to capture images of the gas turbines that power Elon Musk's data center in Memphis. She says the facility is emitting massive amounts of methane - more than some power plants - and calls it "an incredible amount of pollution."

Wilson is outraged that fossil fuels are being burned to generate images of "Nazi Mickey Mouse" and a Grok-like chatbot that spreads conspiracies and misinformation. She calls it a "terrible waste" of resources and harm without public benefit.

Wilson is just one of the experts warning that the artificial intelligence boom comes with a high price, writes GuardianScientists are concerned because it pollutes the natural world with carbon and the digital world with dangers ranging from dubious health myths to fake pornography aimed at children.

Can it Will AI undermine the clean energy transition?

Some experts fear that data centers could derail the transition to a clean economy, adding an unnecessary obstacle to the quixotic task of preventing the planet from warming by 1,5°C (2,7°F).

Others remain more optimistic. They say AI's energy footprint is still smaller than that of polluting industries, and the technology can help develop climate solutions.

But the questions remain: How big a threat is artificial intelligence to the climate? And will it become a climate threat - or part of the solution?

Data center energy crisis comes to the fore

When Hannah Daly worked on models for the International Energy Agency (IEA) in Paris and later became a professor of sustainable energy at University College Cork in Ireland, emissions from digital services seemed insignificant compared to transport, agriculture and heating. But in Ireland, the picture is changing dramatically: data centres already consume around a fifth of the country's electricity and could soon account for a third. Their rapid growth has even led to a temporary ban on grid connections.

Globally, data centers consume just 1% of electricity, but projections point to sharp growth. In the US, their share could reach 8,6% by 2035, according to BloombergNEF, while the IEA predicts that data centers will account for at least 20% of the growth in electricity demand in rich countries by the end of the decade.

Some of this demand is covered through long-term renewable energy purchase agreements - supporting the expansion of clean energy even when the electricity powering the facility is "dirty" - while some tech firms have signed agreements to use nuclear power.

But fossil fuels will remain dominant for the foreseeable future. In China, data centers are concentrated in coal-rich regions, and in the US, where natural gas is expected to generate the majority of data center electricity over the next decade, the Trump administration is using them to justify burning more coal.

"Beautiful, clean coal will be essential to… winning the AI ​​race," Energy Secretary Chris Wright said in September when he announced a $625 million investment package.

AI eats away at climate gains

In Ireland, which is building terminals to import liquefied natural gas (LNG) and power plants to burn it, the data center boom has offset some of the climate benefits of renewable energy sources, according to analysis from last year.

Similar trends are being seen in poorer countries. In Pakistan, cheap solar power is starting to displace coal at a remarkable rate, but data centers are poised to take up the spare capacity of unused power plants after the government announced it would dedicate 2 GW of capacity to artificial intelligence and bitcoin.

"This idea that lower renewable energy prices alone will drive decarbonization is not enough," Daly says. "If there is new demand, it will go to existing fossil fuel capacity."

How much energy does a chatbot use?

Does this mean that using chatbots to write emails, essays and plan vacations is putting the planet at risk? Tech companies have resisted pressure to provide detailed data on the energy footprint used by artificial intelligence, but popular estimates range between 0,2 and 3 watt-hours (Wh) for a simple text query, and increase significantly for "in-depth research" and video production.

In a blog post in July, OpenAI CEO Sam Altman claimed that a request to ChatGPT uses energy comparable to lighting a light bulb for a few minutes. Google reports similar figures for Gemini.

The numbers are insignificant compared to activities like flying, eating meat or driving a car. However, skeptics are concerned about the sheer scale of the technology. ChatGPT claims to have several hundred million weekly users just three years after launching the program - and from the integration of AI into search engines, office software and background services. Google, which holds about 90% of the global search market, has already implemented generative AI into its core product - meaning that each search could require more power than before.

"It's worrying that we're deploying AI in such a way that we don't have a good idea of ​​its actual energy consumption," says Sascha Luccioni, head of climate affairs at AI company Hugging Face.

He criticized large companies for "selective disclosures" that make it difficult to assess the climate footprint of the technology.

"We act as if there is no problem - or if there is, someone else will solve it."

Can it AI to "pay off" its carbon debt?

Some researchers argue that AI can offset its own energy footprint by reducing emissions in other sectors. A report by the International Energy Agency found that existing AI applications can already save more carbon than data centers produce.

An analysis by researchers at the London School of Economics and Systemiq comes to a similar conclusion: AI can accelerate the integration of solar and wind power into the electricity grid, help develop proteins that mimic meat, improve the composition of batteries in electric cars, and encourage more sustainable consumer choices.

"Artificial intelligence can accelerate the deployment of clean technologies by moving them faster along the innovation curve," says Roberta Pierfederici, a policy fellow at the Grantham Institute of the London Foundation for Economics.

However, experts warn that higher efficiency could lead to higher usage - a classic backfire. However, there are already real examples of benefits:

• Google is reducing cooling costs in its data centers by 40%, thanks to AI.
• Spanish company Iberdrola increases the efficiency of wind turbines by 25%.
• French company Engie is shortening the downtime of solar parks through automatic fault detection.

Because other sectors are much more polluting, AI only needs to reduce its own emissions by a small fraction to offset the carbon cost of its calculations, which it says recent study are 0,1-0,2% of global emissions.

"In the energy sector, we are already seeing results," says Pierfederici. "While the meat processing sector is still lagging behind."

When technology accelerates pollution

The transformative potential of artificial intelligence has not escaped the fossil fuel industry, and that's what worries Holly and Will Alpin, two former Microsoft employees who are leaving the company over its growing collaboration with oil giants.

Holly works on the AI ​​sustainability and responsible development teams, and Will was among the first to raise the issue of data center energy costs. But Microsoft's partnerships with ExxonMobil and Chevron-which are accelerating oil production and streamlining drilling operations-are making them question the tech industry's real climate impact.

"The company kept steering the conversation toward its own operational footprint, which wasn't the real problem," says Holly Alpin. After years of internal attempts at change, the two concluded that pressure from within wasn't enough.

The IEA estimates that artificial intelligence could increase oil and gas recovery by 5% and reduce the cost of deepwater offshore projects by 10%. Major oil companies are even more optimistic.

"Artificial intelligence will ultimately be the next fracking boom in the industry," Mike Somers, head of the American Petroleum Institute, told Axios.

Amin Nasser, CEO of Saudi Aramco, reports in an interview with Bloomberg Television earlier this year that the company had implemented artificial intelligence "everywhere," doubling its digitalization spending in just a year. The company said the widespread use of AI had increased productivity.

Satellites see the leaks - but that doesn't stop the emissions

The oil industry claims that artificial intelligence can reduce its carbon footprint, for example by analyzing satellite data to detect methane leaks. But critics like Sharon Wilson stress that monitoring does not lead to real action. She describes "giant clouds of gas" in the Permian Basin and argues that leaks are only a small part of the problem - deliberate methane releases remain widespread.

"Observing methane from space doesn't stop methane," she says.

Even more worrying, some experts say, is the effect on consumption. AI-generated ads are already outperforming human-generated ones, according to a study. study since October, and the ease with which they can be made reduces the cost of promoting consumption. The marketing sector is preparing for AI agents that could buy gifts and book flights on behalf of a customer. Tui, Europe's largest tour operator, says it is investing heavily in artificial intelligence as people turn to ChatGPT to book their holidays.

"It's dangerous to miss the negative use cases," says Holly Alpin.

Some experts are calling for a pause, at least until better regulations are in place. In October, the UN special rapporteur on the human right to safe drinking water called for a moratorium on new data centers, citing the harmful effects on the environment. In December, a coalition of more than 230 environmental groups in the US called for a national pause until the sector is regulated. The Irish Utilities Commission has lifted its de facto ban on grid connections, but has said that 80% of data centers' annual electricity consumption should eventually come from new renewable energy projects.

Attempts to target AI to climate solutions

Others have called for the sector to be incentivized to do good. Spain, the only country to mention AI in climate legislation, obliges the government to promote digitalization, which can help decarbonize the economy. Lawrence Tubiana, one of the architects of the Paris climate agreement, has proposed a tax on AI to fund climate action.

The Alpines, who say they are not against AI but simply want "reasonable safeguards" for the technology, are pushing for the upcoming EU AI bill to classify fossil fuels as a high-risk application of the technology. They also want to force investors to factor in activated emissions in companies' ESG assessments.

OpenAI says it is working to use computing power more efficiently, supports efforts with partners to achieve sustainability goals, and believes artificial intelligence will play an important role in combating climate change.

Microsoft emphasizes that the energy transition is complex and requires "balancing the energy needs and industrial practices of today, while inventing and implementing the innovations of tomorrow."

"AI can help - if we use it wisely"

Sasha Luccioni of Hugging Face believes that instead of panic, there needs to be pressure on companies to create tools that are frugal in design.

"I still believe that AI can help with the climate crisis - designing the next generation of batteries, tracking deforestation, predicting hurricanes," she says. "There are so many good things we could use it for - and instead we're creating websites and social media full of meaningless content while data centers are powered by diesel generators."