365bet�����������籭

ISTANBUL

The EU has long aspired to global leadership in both artificial intelligence and carbon neutrality — two potentially contradicting ambitions that might only be reconciled if the environmental footprint of AI can be massively reduced.

Experts have raised growing concerns about the environmental risks posed by AI technologies and the data centers that support them, forcing Europe into a balancing act between innovation and sustainability.

The UN Environment Programme (UNEP) has warned of AI’s mounting environmental toll — from e-waste and massive water use for cooling servers to high electricity consumption and reliance on polluting rare minerals.

Europe is aiming to become a global leader in AI, but also wants to be a green continent

Can these two things really go together? ⬇️

— Anadolu English (@anadoluagency)

To put that into perspective, manufacturing a single two-kilogram computer requires extracting around 800 kg (over 1,760 pounds) of raw materials, according to the UN’s 2024 Digital Economy Report.

One estimate cited by the UNEP suggests that annually, AI infrastructure could soon consume six times more water than Denmark. The International Energy Agency expects the AI industry to consume at least 10 times more electricity by 2026 than it did in 2023.

Beyond international bodies, others are also sounding the alarm. Shaolei Ren, an associate professor of electrical and computer engineering, and Andrew Winston, founder of Winston Eco-Strategies, both emphasized AI’s growing environmental footprint.

“Compared to other type of computing workloads, AI’s environmental footprint is definitely increasing, and at much faster speed. Probably, we have never seen such growth rates before for other workloads,” said Ren of the University of California, Riverside. He warned that building more data centers is “not good news” for the climate.

Winston highlights that all technologies and data centers carry a significant footprint, with AI now driving almost all the sector’s growth. “The emissions are rising. The energy use is rising faster than the gains.”

Winston also pointed to recent environmental reports from Microsoft and Google, showing that emissions are increasing even as both companies had pledged reductions. Microsoft’s 2024 report revealed a roughly 30% increase in indirect carbon emissions, mostly from data center construction. Meanwhile, Google reported a 48% jump in greenhouse gas emissions in 2023 compared to its 2019 baseline.

“When people build AI systems, they want more stable energy,” Ren explained. “That could be a problem for this renewable energy adoption, because typically, the more renewables you have in your grid, the more intermittent energy supply you have — and that is not something data centers want to have.” He noted that some countries are already building natural gas power plants for this reason.

Environmental risks go beyond emissions

Winston stressed that AI could be a powerful tool to combat climate change by helping companies operate more efficiently — but warned it could also be deployed to accelerate environmentally harmful activities, such as locating new fossil fuel reserves.

“It depends on what we’re pointing AI at,” he said.

“Are we generally using it to reduce footprint, or are we figuring out ways to find more fossil fuels, ways to make fast fashion even faster, and send apparel around the globe? It just depends on what we're using it for. There are things that make emissions worse, and we're going to get good at those, too.”

Experts call for a holistic approach

Ren said it is crucial to look at AI’s environmental impact holistically, rather than focusing solely on carbon emissions.

“We can estimate the amount of carbon emissions by certain amount of a certain type of AI models. However, the downstream impact of this amount of carbon emission on the global climate is unknown,” he said. “There’s a lot of uncertainty there.” he noted, underlining the need to look at more than just raw emission while evaluating AI’s impacts.

He noted that EU directives requiring companies to report energy, water, and carbon usage are helpful, but limited.

“It’s not trying to cure industry. It’s trying to make sure that we have a bottom line for this industry development. So, in the end, it depends pretty much on how the industry evolves,” he added.

Winston said companies need to ask “deeper” questions about the purpose of their AI systems.

“If we’re not heading towards solving the world's problems through our business, then we're causing destruction, and we’re making it harder for everybody to live a thriving life,” he said. “I hope the AI companies kind of consider the power that they're wielding and point it in the right direction.”

He added that while more efficient chips and data center designs can help, the rapid expansion of AI infrastructure still demands large-scale energy investments, which could either accelerate or delay the global transition to clean power.

Time running out for climate action

Winston warned that the world is running out of time to address climate change, and that choices made now — including how AI is developed — will have long-lasting consequences.

“I think this is a fundamental choice now about, are you building a better world or not? Because we’re really running out of time. The problems we face are intense. Climate change is getting worse rapidly,” he stressed.

He called AI the “most important technological change since the Internet” — one that is evolving quickly and in ways we do not yet fully control.

“We’re letting it decide its own fate in many ways,” Winston said. “I hope they’re programming in not just moral responsibility, but environmental responsibility, too.”