As AI models become more complex, their environmental, resource, and economic impacts are growing. Organizations are aiming to meet the moment with standardized frameworks and coalition-building.

Generative AI has rapidly shown the ability to improve business operations, delivering greater productivity and performance. But as companies reap the benefits of these new tools, many are confronting another reality: AI’s steep environmental cost. 

Powering AI at scale requires an enormous amount of electricity, with large Gen AI models consuming up to 4,600 times more energy than traditional models. And the demand for more power and AI compute is only rising. In a high-adoption scenario with many companies deploying complex AI models, such as agentic AI, electricity use is projected to rise by a factor of 24.4.1 This would lead to higher carbon emissions, strains on resources like water, and negative community impacts. Temperatures in neighborhoods around data centers are generally higher by 2°C than equivalent neighborhoods without this infrastructure.  

For businesses adopting AI, this prompts a key question: how can I deploy AI at scale, while mitigating its environmental impact? 

Driving value through internal collaboration 

Companies must seek to align on AI across their internal teams, as perspectives can vary widely across company functions and AI-related decisions are often made in silos. Consequently, organizations may optimize their AI use for certain objectives, while unintentionally making it harder to achieve others.  

To scale AI sustainably, organizations need clear communication and a common shared framework based on standardized metrics covering environmental, resource, and economic impacts and objectives. Creating this shared language enables effective decision-making across all aspects of a business. 

Take transport and logistics. AI is being increasingly used in this sector to optimize operations, but its value depends on where, when, and how it is deployed, as competing priorities must be balanced. Business teams may focus on reducing delivery times and lowering fuels costs, while sustainability teams want to cut emissions, and technology teams prioritize model performance. To optimize these different objectives, a common measurement framework can help quantify value by considering costs, resource use, environmental impact, and increased efficiency. 

Creating industry-wide alignment 

While action at company level is vital to managing the relationship between business value, resource consumption, cost, and environmental impact, industry-wide standards are also critical. Isolated measures that address hardware efficiency, model efficiency, or grid improvements alone cannot mitigate Gen AI’s impact. Achieving this will require coordination across the AI value chain, from power generation to operational efficiency, to the product lifecycle. 

Governments and businesses are already stepping in to minimize AI impact and align AI development with net-zero goals. The International Telecommunication Union (ITU) recently published guidelines for assessing the environmental impact of AI systems (ITU-T L.1801) and other efforts are underway. One French-led initiative, the Coalition for Sustainable AI, brings together the UN Environment Programme, ITU, Capgemini, and many other companies and organizations to address AI’s impact. Coalition members are committed to reaching the UN Sustainable Development Goals and UN Agenda 30, leveraging AI tools to support climate action and environmental protection. 

  1. Clément Desroches et al., “Exploring the sustainable scaling of AI dilemma: A projective study of corporations’ AI environmental impacts,” Capgemini Invent, https://arxiv.org/pdf/2501.14334