Key facts
- Data centers and AI are not on the agenda for the upcoming UN climate talks in Turkey.
- The International Energy Agency estimates data center climate pollution could double by 2035 to 350 million metric tons of greenhouse gases.
- Proposed data centers in the U.S. could lead to new natural gas plants emitting as much carbon as Australia.
- Emissions from planned diesel generators for two UK data centers could exceed ExxonMobil's operations in the country.
- The UN's AI for Climate Action initiative, launched in 2023, examines AI's role in climate solutions and risks.
- Phoenix Mayor Kate Gallego called for elevating the issue of data centers in the international arena.
The United Nations climate talks, set to convene in Turkey in November, are facing criticism for not including the rapidly growing energy demands of data centers and artificial intelligence on their formal agenda. This oversight comes as AI development fuels a surge in electricity consumption, potentially leading to increased reliance on fossil fuels and accelerated greenhouse gas emissions.
According to a COP31 official, the issue has not been formally proposed for inclusion by any country. While the UN has addressed data centers' impact once, in a paragraph within the 2023 Dubai talks' outcome agreement that noted AI's potential environmental challenges, critics argue the UN's bureaucratic pace is too slow to match the speed of technological advancement.
Analysts estimate that data center climate pollution could double by 2035 to approximately 350 million metric tons of greenhouse gases, a figure comparable to the UK's annual emissions. In the U.S., this growth is driving a natural gas boom, with proposed data centers potentially leading to carbon emissions equivalent to Australia's. Some European data centers are opting for renewable energy but are also installing backup diesel generators, and emissions from planned UK data centers could surpass ExxonMobil's local operations.
Outside the formal negotiations, the UN has initiated efforts to address AI's role in climate solutions and its risks. UN Secretary-General António Guterres has urged AI companies to disclose their environmental impacts and commit to clean energy by 2030. However, integrating new issues into consensus-driven UN talks is a lengthy process. Some advocates suggest data center concerns could be addressed through voluntary initiatives like the COP's action agenda, which prioritizes electrification and aims for 35% of global energy use to be electricity by 2035, supported by cleaner power, stronger grids, and energy efficiency.
Big tech companies and climate diplomats highlight AI's potential to improve grid efficiency and renewable energy use, potentially offsetting its emissions. However, this also risks widening global disparities. Some officials and advocates believe data center issues can be managed outside formal negotiations through panels on grid management, electrification, or energy efficiency. Kaveh Guilanpour of the Center for Climate and Energy Solutions noted that increased energy use for data centers is not inherently contradictory to climate goals, provided emission savings are found elsewhere in the economy.
Despite the slow pace, the UN is beginning to elevate the issue. The UN's academic arm warned in June about AI's sweeping environmental impacts and the risks of compromising environmental governance due to the rush for data centers. UN climate chief Simon Stiell has called for greater energy efficiency and clean power, stating that data centers should not lead to wider development gaps or higher emissions. The AI for Climate Action initiative focuses on how AI can help developing countries with climate challenges. Phoenix Mayor Kate Gallego advocates for elevating the data center issue internationally, though she suspects other venues may be more impactful in the short term. Small island nations, represented by Anne Rasmussen of AOSIS, are concerned about any significant new energy demand, emphasizing that technological progress must align with rapid decarbonization.