According to a recent report by the International Energy Agency (IEA), global data centres—the digital engines of AI—are set to consume a staggering 945 terawatt-hours of electricity annually by 2030. That's more than an entire country like Japan uses in a year.
The sheer power required to train and operate advanced AI systems, such as ChatGPT, is immense. A single training session for such a model can use over 1,200 megawatt-hours of electricity—emitting as much carbon as dozens of short-haul European flights. Though responding to individual queries uses less energy, the numbers quickly add up with over a billion AI interactions now happening daily.
This rising energy demand is concentrated in the US, which hosts roughly 40% of the world's data centres. With most of these still running on fossil fuels—especially natural gas—climate experts are raising red flags. The IEA warns that without decisive policy shifts, the US is likely to experience the sharpest increase in AI-driven energy use.
Recent political developments suggest the trend may continue. In early 2025, President Donald Trump signed an executive order directing officials to explore coal-powered infrastructure options for new AI data centres. In Louisiana, for instance, a gas-fired power plant is already being built to support a massive Meta facility.
Despite these concerns, the IEA maintains that AI has significant potential to drive down emissions in other sectors. Used strategically, AI could optimise industrial efficiency, streamline power grids, and boost the integration of renewable energy. The agency estimates that, if widely adopted, AI could help cut global emissions by up to 5% by 2035—potentially offsetting its own carbon footprint.
However, this positive outlook is far from certain. The IEA cautions that there is currently no strong momentum behind widespread climate-friendly applications of AI. Unless investment and regulation align, its net impact could be negligible—or even harmful.
Some tech giants are leading by example. Amazon, for instance, claims that more than 90% of its operations—including its AWS data centres—are powered by renewable energy. Data infrastructure company Digital Realty reports that its entire European portfolio is now matched with green electricity. Yet challenges remain, including the inconsistent availability of clean energy and the limits of current grid technology.
AI's future role in climate action will depend on whether its growth is matched by meaningful environmental responsibility. The IEA closes its report with a clear message: AI might help reduce emissions in theory, but the scale of change required to tackle the climate crisis still far exceeds what AI can offer alone.