Artificial intelligence has entered the heart of climate negotiations. THE November 16, 2026during the COP31 organized at Antalyain Türkiye, the Turkish presidency of the conference and theInternational Telecommunications Union (ITU), the United Nations digital agency, launched theAntalya commitment on artificial intelligence. This voluntary text aims to both put AI at the service of the climate and limit its environmental footprint. But its launch above all highlighted those who were absent.
“Immense potential and great danger”
The tone was set by Simon Stiellexecutive secretary of the UN Climate. “We have arrived at a technological moment of both immense potential and great peril. AI and other advanced technologies are a double-edged sword,” he said at the launch. “We cannot ask people to accept higher bills, more planet-warming pollution and deeper inequality now, in exchange for promises of a better future. The benefits of AI must outweigh its growing costs. »
The UN official pointed out the demand for data centerswhich energizes the electrical networks. “Where fossil fuels meet this growing demand, pollution increases and our climate goals slip away,” he added, noting that some developers are building gas power plants on their own sites to bypass network limitations. He also recalled that residents are already contesting data center projects, worried about their water, their bills and their environment.
What the commitment provides
According to Balkan Green Energy News, Antalya’s commitment is structured around four main axes:
| Axis | Content |
|---|---|
| Sustainable infrastructure | Power and build AI infrastructure and hardware sustainably, with improved energy and water efficiency |
| Responsible design and purchasing | Design, manage and purchase AI responsibly, with a circular approach to waste |
| Measurement of impacts | Improving the assessment and publication of the environmental impacts of AI |
| Support for developing countries | Strengthen financing and technical infrastructure for AI for the climate |
The signatories undertake to act according to international standards for AI systems, based on electricity low carbon or no emissions. The text, however, remains voluntary: it sets no quantified objective, no binding timetable, no sanction.
Support still very limited
At its launch, the pledge garnered 80 supportersmainly coming from businesses and civil society organizations. Alone nine countries formally signed it. L’Germany announced that she would support it, but the UNITED STATESthere China and theEuropean Union are not part of it.
On the business side, the main players in AI are also missing: Nvidia, OpenAI And Anthropic are not on the list. THE Global Business Council for Sustainable Developmentwhich counts Amazon, Google and Meta among its members, has however signed it.
“Eighty is not a limit,” Tomas Lamanauskas, deputy secretary general of the ITU, assured AFP, who sees this launch as a starting point intended to create momentum.
Criticisms about a “green label”
Environmental organizations remain skeptical. Jill McArdle, of the Beyond Fossil Fuels association, calls the commitment “a diversion” from “vague promises”. “It risks offering tech giants a green label without forcing them to tackle the real climate impacts of their infrastructure,” she believes.
The criticism focuses on a central point: without quantified objectives or obligation of transparency, there is no guarantee that the signatories will actually reduce the footprint of their data centers. In June 2026, UN Secretary-General António Guterres had already called on AI companies to publish their environmental impact and power all their data centers with renewable energy by 2030.
The AI footprint in numbers
The figures cited during the conference give an idea of what is at stake:
- according to theInternational Energy AgencyTHE coal provides approximately 30% electricity consumed by data centers, ahead of renewables (27%), gas (26%) and nuclear (15%);
- according to a UN study cited by AFP, data centers will consume in 2025 more electricity than all but ten countries in the world;
- by 2030, their water consumption could cover the annual basic needs of the entire population of sub-Saharan Africa, according to the same study;
- In the United States, the most powerful data centers accounted for approximately 2% of electricity consumption, according to a study published in Nature Sustainability.
This dependence on fossil fuels is observed in recent projects: Anthropic has secured gas generators to power its data centers, and SpaceX is planning a gas power plant for its future computing campus in Texas.
In parallel with the commitment, Turkey opened its AI-based clean technology initiative internationally, led by the GreenWise Impact Investment fund. It connects climate challenges, AI start-ups and investors, in order to accelerate the deployment of solutions in the industry.
And France?
France was not cited among the first signatories in the launch reports. However, it has a strong argument in this debate: very largely carbon-free electricity thanks to nuclear and renewables, put forward to attract data center projects. The absence of the European Union, which already has rules on the energy efficiency of data centers, shows, however, that the major blocs prefer their own frameworks to a voluntary international commitment.
What businesses can already do
Without waiting for international rules, organizations that use AI have simple levers to limit their footprint. Choose it model adapted to the task rather than the most powerful greatly reduces the necessary calculation: a small model is often enough to summarize or classify documents. Favor data centers located in regionslow carbon electricity and asking your suppliers about the origin of their energy and their water consumption are two other useful reflexes. Finally, measuring the actual use of AI tools helps avoid oversized deployments.
A debate set to intensify
COP31 continues until November 20, 2026, under the Turkish presidency and with Australia leading the negotiations. Antalya’s commitment alone will not change AI’s energy trajectory, but it puts the issue on the global climate agenda. With the proliferation of giant computing campuses and pressure on electricity grids, the question is no longer whether AI must be accountable for its footprint, but how and according to what rules.

