Sopra Steria unveils its report "AI & the Environment: clearing the information fog" - an unprecedented analysis of narrative biases, heterogeneous measurement methods and new management tools - and calls for a sober and transparent approach to regulate the entire life cycle of AI.
Although the massive adoption of generative AI is considerably amplifying the environmental footprint of digital technology, this subject remains largely under-discussed: only 0.56% of online conversations and 2.4% of media articles concerning AI mention its link with the climate.
Beyond this observation, Sopra Steria advocates a third way: neither naïve solutionism nor ideological rejection, but sober, reasoned AI designed to last.
Paris, 26 June 2025 - Faced with the exponential adoption of generative artificial intelligence, Sopra Steria, a major player in European tech and a trusted partner for ethical AI, today publishes its second report dedicated to the fight against information: "AI & Environment: clearing the information fog". After a first opus devoted to misinformation, this new report unveils a part that is often neglected: the real environmental impact of AI.
The environmental footprint of AI, a public and media blind spot
At a time when global data centre consumption could almost triple by 2030 (from 415 TWh in 2024 to over 940 TWh according to the IEA), the rise of AI is amplifying an already critical energy dynamic. Technology giants such as Microsoft and Google have seen their CO₂ emissions jump by +30% and +48% respectively in recent years, illustrating the environmental cost of poorly controlled innovation.
Yet this cost remains surprisingly absent from the debates. Sopra Steria, drawing on the social listening expertise of Opsci.ai, analysed more than 800,000 messages published on X and LinkedIn between January 2024 and April 2025: only 0.56% mention the ecological impact of AI - three times less than for cryptocurrency. As for the media, out of nearly 22,000 articles on AI, barely 2.4% focus on its environmental impact, often with a techno-solutionism narrative that plays down the risks. Even the AI leaders identified by Time Magazine do not consider climate a priority - on analysis of their publications, the subject only comes in 8ᵉ position, well behind innovation or productivity.
However, a detailed analysis of the data and conversations shows a growing polarisation in the debate: technophiles promote AI as a universal solution, while technophobes reject all progress out of hand.
Between these two extremes, the scientific community, which is essential for objectivising the debate, often finds itself instrumentalised or invisible, a sign of wider distrust of scientific expertise in the public arena, the politicisation of debates and the amplification of misinformation.
Axelle Lemaire, Executive Director of Sustainable Performance at Sopra Steria, says: "Who hasn't fallen for the 'magic' of AI? However, this study highlights a worrying paradox: while the environmental footprint of artificial intelligence is increasing, the correlation between AI and climate change remains largely absent from public and media discourse. This invisibility is amplified by the logic at work around climate disinformation, mistrust of science and ideological polarisation, not to mention the huge financial stakes involved. All of this contributes to maintaining opacity, preventing lucid debate and taking us further away from the objectives set by the Paris Agreement. We urgently need to get out of this fog of information about the impact of AI and demand transparent measurement methods and standards, so that more sober, useful and reasonable uses emerge, in line with the climate challenges of our time".
Clearing the fog: the need for common tools and reliable indicators
At a time when the mass adoption of AI and generative AI is intensifying, this relative invisibility is compounded by a crying lack of standardised methods for assessing and monitoring the environmental impact of digital technologies. This methodological vagueness hinders the adoption of more sober solutions, complicates the implementation of common standards and makes it difficult to be transparent with customers and regulators.
In this context, Sopra Steria is therefore calling for the alignment of measurement standards at European level and the widespread use of transparent, verifiable and comparable indicators, to integrate digital sobriety into the design of technologies.
Taking concrete action: a collective commitment to sustainable AI
Today, the lack of harmonised standards hinders objective comparisons between models and delays the adoption of more sober solutions. In response, Sopra Steria is actively supporting the construction of common benchmarks, supported by concrete regulatory and sector initiatives: the REEN Act, a pioneering framework for data centre transparency; the European AI Act, which imposes environmental reporting obligations for high-risk AI; and the Coalition for Sustainable AI, a group of players committed to more responsible and measurable artificial intelligence.
At the same time, the Group is developing practical tools: EcoMind AI, to anticipate the environmental footprint of AI models right from the design stage, and G4IT to quantify the overall carbon impact of information systems. These systems are part of an impact ROI strategy, which integrates social and environmental benefits into the assessment of AI projects.
Sopra Steria defends a balanced approach: refusing the alternative between a naïve solutionism and a pure and simple rejection of technology. The company is committed to AI that is useful, measurable and responsible.
Yves Nicolas, Director of the Sopra Steria Group's AI programme, concludes: "Artificial intelligence can no longer be thought of in isolation from its impacts, including environmental impacts. The development of responsible AI should not be an option, but a collective requirement. At Sopra Steria, we are convinced that it is by integrating sobriety criteria right from the design phase that we will be able to reconcile technological innovation and ecological transition. This requires reliable indicators, greater transparency and constant dialogue with all our stakeholders - three requirements that are part of our AI roadmap".