International Green AI inititiatives
There is currently a need to understand the environmental costs of AI to ensure deployment at scale for all countries and to build an AI ecosystem respectful of our planetary boundaries. The goal of the Coalition is to ensure that AI becomes an essential lever for addressing global environmental challenges while minimizing its ecological footprint. Various initiatives will enable this vision, starting with methods and standards to measure and minimize this impact along the value chain.
To ensure that the use and development of AI is in line with planetary boundaries, this inventory of initiatives brings visibility to Green AI collaborative projects that aim to reduce or minimize the environmental impacts of AI.
These projects were selected following consultation with members of the Coalition for Sustainable AI. They are organised into sub-categories according to how they address AI environmental issues.
Standards
Roadmap for AI environmental sustainability standardization
The objective of this roadmap is to ensure efficient use of resources, reduce confusion, promote consistency in the measurement of the environmental impact of Artificial Intelligence (AI), and facilitate the widespread adoption of best practices in that regard. Contributors wish to work towards non-conflictual standards to measure the environmental impact of AI and encourage collaboration between international standardization bodies to avoid, as far as possible, the duplication and overlaps of standards.
Environmental Impacts of Artificial Intelligence Working Group
This standard defines a measurement framework for reporting on environmental indicators for training models and deriving inference on Artificial Intelligence (AI) systems. This includes harmonized measurements of compute intensity (e.g. energy use) with associated environmental impacts (e.g. carbon dioxide (CO2) emissions, or water consumption). This standard describes methodologies to separate the measurement of AI-specific compute (i.e. data centres used for AI training or inference) from general purpose compute (e.g. data centres used for other purposes like cloud services).
Metrics and Guidelines for the Environmental Impact of AI
This document describes the principles and framework for environmental impact measurement of artificial intelligence systems and services and provides guidelines for impact reduction throughout the lifecycle. It includes a framework for defining the environmental impact of artificial intelligence, aharmonized calculation method for assessing the environmental impact of artificial intelligence systems and services, reporting guidelines as well as best practices for reducing the environmental impact of AI systems and services throughout their lifecycle.
Guidelines for assessing the environmental impact of AI Systems
This report aims to fill this gap based on new global and regional modelling and datasets, as well as extensive consultation with governments and regulators, the tech sector, the energy industry and international experts. It includes projections for how much electricity AI could consume over the next decade, as well as which energy sources are set to help meet it. It also analyses what the uptake of AI could mean for energy security, emissions, innovation and affordability.
The iMasons Climate Accord
The iMasons Climate Accord coalition works towards the industry adoption of an open standard to report carbon in data center power, materials, and equipment and a maturity model to report participant progress.
Software Carbon Intensity for AI
The purpose of this proposed specification is to assist AI practitioners in understanding and reducing the carbon footprint of AI systems. By making informed choices about model design, computational efficiency, and deployment strategies, practitioners can minimize emissions while maintaining performance. A working group is developing a methodology for calculating the carbon emissions rate (SCI score) of AI software systems, including both classical AI and generative AI applications. This specification aims to provide a reliable, consistent, and comparable measure that practitioners can use to set targets and track progress in reducing carbon emissions throughout the AI lifecycle.
Best practices
Engagement Kit for Frugal AI
The French Ministry of Environment, in partnership with the Hub France IA and AFNOR Standardisation, is publishing and submitting for consultation a list of best practices on frugal AI for public and private stakeholders. Each organisation can adapt this list and cite it in AI charters, such as ethics charters. This list aims to encourage organisations that adopt it, to have a reasoned development of the technology, by questioning the need for AI above all and by promoting efficient AI systems. This list of best practices aims to propose operational and achievable actions.
Sustainable Datacenters & ICT
The project is led by the Copenhagen Centre on Energy Efficiency and serves as a global platform to share best practices of energy efficient solutions in data centres, provide policy recommendations to policy makers and foster transformational collaboration among various public and private stakeholders in the global data centre industry.
Library of compression algorithms
Pruna is a model optimization framework built for developers, enabling you to deliver faster, more efficient models with minimal overhead. It provides a comprehensive suite of compression algorithms including caching, quantization, pruning, distillation and compilation techniques
Climate Neutral Data Centres Pact
Over 100 data centre operators and trade associations are committed to the European Green Deal, achieving the ambitious greenhouse gas reductions of the climate law, and leveraging technology and digitalization to achieve the goal of making Europe climate neutral by 2050. To ensure data centres are an integral part of the sustainable future of Europe, data centre operators and trade associations agree to make data centres climate neutral by 2030.
Measuring & Reporting tools
Code Carbon
CodeCarbon is a lightweight software package that integrates into a Python codebase. It estimates the amount of carbon dioxide (CO2) produced by the cloud or personal computing resources used to execute the code. It then shows developers how they can lessen emissions by optimizing their code or by hosting their cloud infrastructure in geographical regions that use renewable energy sources
Green Algorithms
The Green Algorithms project aims at promoting more environmentally sustainable computational science. It regroups calculators that researchers can use to estimate the carbon footprint of their projects, tips on how to be more environmentally friendly, training material, past talks etc.
Ecologits
EcoLogits tracks the energy consumption and environmental impacts of using generative AI models through APIs. The main goal of the EcoLogits Calculator is to raise awareness on the environmental impacts of LLM inference.
CarbonAI
This project aims at creating a python package that allows you to monitor the power consumption of any python function.
Carbon Footprint Calculator for AI
AI carbon footprint calculator allows to easily assess the environmental impact of a business’s AI marketing activities.
Net Carbon Impact Assessment Methodology for ICT Solutions and case Study Calculators
The European Green Digital Coalition developed science-based methods to estimate the net environmental impact of real-life digital solutions across sectors – Energy/Power, Transport, Construction/Buildings, Smart cities, Manufacturing, Agriculture. The main outcomes of the EGDC Phase I were a science-based method to calculate the net climate impact of given solutions, Case Study Calculators, which contributed to refining the Net Carbon Impact Assessment Methodology, and deployment guidelines to provide recommendations for green digital transformation in each of the six sectors.
BoAmps
This project is composed of a standard datamodel (with tools to fill it easily) and an open data space for collecting and sharing energy consumption measures from AI systems. The goal is to shed light on the materiality (hardware and power consumption) induced by the use of AI, with the aim of deducing the environmental impacts associated with their use, with a systemic approach. This database could then be used to convert electricity and resources into a multi-criteria environmental impact, based on user-specific references to enable LCAs to be carried out.
Energy and Environmental Optimization of Data Centers
The research project aims to analyze and anticipate the growing global impact of data centers on energy consumption and CO₂ emissions, to reconcile digital development with sustainability. The project comprises four research axes: development of a decision-support dashboard to guide siting or technological transformation; modeling the thermal sensitivity of data centers by cooling architecture; global mapping (present and future) of energy and carbon footprints; energy optimization of computing workload distribution across interconnected hyperscale data centers; development of a decision-support dashboard to guide siting or technological transformation.
Transparency
Energy & AI Observatory
Faced with the growing challenges related to the energy consumption of Artificial Intelligence and the opportunities it represents for the energy sector, the International Energy Agency is launching a Global Observatory dedicated to Energy and AI. This initiative, carried out in collaboration with industry players and scientific experts, aims to provide a global and informed vision of the impact of AI on the energy sector. It provides up-to-date data and analysis on the growing links between the energy sector and artificial intelligence.
Observatory of AI Environmental Impact
With the support of Capgemini, the AI and Society Institute, the Ecole normale supérieure (ENS-PLS) and the ENS Foundation have launched an Observatory dedicated to analyzing and mitigating the environmental impacts of Artificial Intelligence (AI) at all stages of its lifecycle (training, adjustment, inference and end-of-life). The new Observatory aims to establish a solid, shared methodology to encourage sustainable AI usage.
Expert Group on AI Compute & Climate Change
The OECD.AI Expert Group on AI Compute and Climate contributes to the OECD’s initiative to create a basic framework for understanding, measuring and benchmarking domestic AI computing capacity by country and region. While mindful of the ever-evolving state of the computing landscape, the expert group is working with key AI computing players in a data-gathering exercise to understand “ai compute” energy consumption.
Green Digital Action Working Group on Sustainable AI
The Green Computing pillar of the Green Digital Action (GDA) initiative, under the auspices of the International Telecommunication Union (ITU) together with over 50 partners, has embarked on a comprehensive three-phased project to tackle the challenge of the absence of clear, standardized metrics for measuring AI’s environmental footprint. The first phase involves an overview and meta-analysis of current efforts to measure the environmental impact of AI. In the second phase, the project will develop a comprehensive measurement and testing plan. Finally, the third phase will implement this plan, assessing the environmental impact of various AI scenarios and workloads.
AI Energy Score
The AI Energy Score initiative, co-led by Hugging Face and Salesforce in partnership with Cohere and Carnegie Mellon University, aims to establish a standardized framework for evaluating the energy efficiency of AI model inference. This project addresses the urgent need to assess and mitigate the environmental impact of AI systems, which are projected to consume significant amounts of energy in the coming years. This initiative needs the community’s help to help the project realize its full potential (more information).
Impact AI
This research project examines AI projects oriented towards the public interest with a focus on sustainability and develops a transdisciplinary auditing method to evaluate their contribution to societal transformation and ecological sustainability. The goal is not only to make the social impact of AI systems measurable but also to identify concrete actions for their responsible use. To bridge scientific analysis with practical applicability, the evaluation method developed will be made publicly available. This ensures that both trained experts and organisations can directly foster the auditing method. In this way, Impact AI contributes to strengthening the sustainable use of AI systems and enabling informed decisions for technologies that serve the public interest.
