This November, COP30 convened in Belém, Brazil — marking a historic milestone as the first COP summit to place artificial intelligence (AI) and data-driven technologies at the heart of its agenda. While traditional disputes, such as those surrounding fossil fuel use, once again stalled negotiations over the final political agreement, AI emerged as a rare unifying theme. Top officials — including COP30 President André Aranha Corrêa do Lago and UNFCCC Executive Secretary Simon Stiell — publicly emphasized the vital role of digital tools in tackling climate change.
AI applications surged in 2025, with a growing number of sectors exploring the technology’s transformative potential. While AI has been in use for over a decade in areas like scientific research and medicine, the energy sector has been slower to adopt it. “There is definitely room for improvement,” said Laura Cozzi, Director for Sustainability, Technology, and Outlooks at the International Energy Agency (IEA).
AI excels at processing vast datasets and identifying patterns, making it especially promising for optimizing energy use. “For instance, simply optimizing traffic routes could equate to removing 120 million cars from the roads,” Cozzi noted.
Indeed, AI is already supporting climate efforts in meaningful ways. Many applications leverage image recognition and data analysis. The Ocean Cleanup project uses AI to detect plastic waste in the oceans and guide cleanup operations. Other projects track deforestation via satellite imagery or help scientists monitor changes in polar ice.
Despite these benefits, public debate has zeroed in on the energy consumption of AI technologies — particularly data centers. A widely circulated claim suggests ChatGPT consumes ten times more energy per search than Google. This, coupled with the substantial water needed to cool data centers, has fueled concerns that AI could become a Pandora’s box for emissions.
Yet the reality is more complex. According to the IEA’s April 2025 report, Energy and AI, data centers currently account for only 1.5% of global energy consumption — with just 25% of that tied directly to AI. However, these facilities place disproportionate strain on local power grids due to their need for high-quality, uninterrupted electricity. To ensure stable supplies, the IEA recommends — and permits — the use of natural gas and nuclear energy.
Another challenge is the misalignment between the timelines for data center construction and grid development. “It takes just one to two years to build a data center, but eight to ten years to construct a supporting grid,” Cozzi explained. This highlights the need for stronger collaboration between data centers and energy providers.
Copyright: Canva. AI meets Nature
However, the demand for new grids may be overstated, suggests Amory Lovins, a leading American energy policy expert. In his essay, Artificial Intelligence Meets Natural Stupidity: Managing the Risks, Lovins argues that data centers are far more flexible in their energy use than commonly assumed. He estimates they account for 25% of U.S. electricity demand growth, compared to the U.S. Energy Information Administration’s estimate of 62%. Lovins advocates for a more nuanced and cautious approach to energy policy regarding data center expansion.
A critical — and often overlooked — question remains: How much greenhouse gas (GHG) emission is actually driven by AI and digital technologies? According to the IEA, emissions are estimated to range between 300 to 500 million tons (0.9–1.5% of total global emissions) during the so-called “lift-off” phase, when new data centers are rapidly constructed. “Is it going to make or break climate change? No,” Cozzi stated. “But it is undoubtedly a polluting sector, and its impact is growing — so caution is necessary.” Lovins further warns that AI’s efficiency at locating fossil fuel reserves might undermine its climate benefits if it leads to more oil and gas production.
Recognizing these challenges, COP30 called for increased international cooperation in AI development and deployment — particularly in relation to climate action.
One standout initiative launched in Belém was the Green Digital Action Hub, a global digital platform that promotes best practices in digital transformation and fosters collaboration. A key focus is on equipping the developing world with practical digital tools. Notably, although 50% of global internet demand now comes from countries outside the dominant digital infrastructure hubs (North America, Europe, and China), only about 10% of data centers are located in these regions.
In support of this goal, the AI Climate Institute was also established to help Global South countries adopt AI-based climate solutions. Several other COP30 initiatives referenced AI and digital cooperation, such as the Tropical Forest Forever Facility, which provides financial incentives for rainforest conservation, and the global mutirão, an ambitious plan to triple climate adaptation funding by 2035.
Overall, the energy sector has officially welcomed AI — with all its promises and perils. Although the first regulatory steps have been taken, the global landscape remains largely uncharted. Are we on the verge of another tech bubble and collapse? Perhaps. But this time, we might be better equipped to see it coming.