Developing sustainable GenAI

Summary

A white paper published by Capgemini in early 2025 describes the environmental impacts of generative AI, the lack of corporate awareness regarding this footprint, and proposes a roadmap to mitigate the impact.

The study covers the entire life cycle of AI, including various criteria such as energy consumption, water usage, and abiotic resources.

It provides an in-depth analysis of why companies fail to consider the environmental footprint of generative AI and outlines the key factors contributing to this issue.

The study highlights the lack of awareness about AI’s environmental impact, supported by testimonials and data from industry professionals.

In its final section, the study presents a roadmap for making generative AI applications more sustainable. Capgemini explores various strategies, starting with measurement, and proposes technical approaches to better manage AI’s environmental impact.

The editorial perspective

From the outset, the title is striking and somewhat bold considering the environmental footprint of generative AI. However, it effectively sets the framework for this white paper.

With great precision, Capgemini thoroughly describes the environmental impact of AI across its entire life cycle. Each description is presented in a clear and educational manner, with diagrams and tables supplementing the study’s conclusions throughout. It is easy to read and understand.

The study highlights the lack of awareness about generative AI’s environmental footprint within organizations. However, it also notes that business decision-makers are increasingly aware that generative AI contributes to rising corporate greenhouse gas emissions (47%). In this context, 42% of organizations that have started using generative AI acknowledge that they are revising their initial ecological transition commitments and goals.

The study also emphasizes that AI’s environmental impact remains a low priority for most companies. Another key takeaway is the detailed examination of research efforts by major generative AI firms to reduce AI’s impact.

In the final section, Capgemini presents a roadmap for "sustainable Gen AI." According to Capgemini, implementing a sustainable approach to generative AI is undoubtedly feasible. However, introducing alternative perspectives and a systemic approach would have added more nuance to the roadmap. As currently structured, this final section could have fostered a broader debate on what choices should be made to prevent AI from becoming a catch-all solution for organizations' challenges.

That said, the roadmap’s consistency and technical and strategic proposals are commendable. Once again, Capgemini examines the entire life cycle of Gen AI projects, highlights relevant use cases, and effectively explores governance issues.

Overall, the report is a solid tool for IT decision-makers (DSI) to drive greater consideration of AI’s environmental impact within their organizations and work toward reducing its footprint.