How Capgemini's Collaboration Will Tackle AI Sustainability

AI has a sustainability problem. Widespread adoption of the technology, especially Gen AI, has unlocked undeniable opportunities across industry sectors worldwide. 

But the growing computational demand needed to meet new AI use cases and applications is increasing its environmental impact in terms of energy and water consumption, as well as carbon footprint.

To analyse and mitigate these challenges the Ecole normale supérieure (ENS-PSL), AI & Society Institute and Capgemini have launched a global observatory that aims to establish a robust, shared methodology to spearhead sustainable AI development.

“Companies are looking for efficiency gains, yet very large AI models not only entail a significant environmental footprint, but also higher costs,” says Etienne Grass, Managing Director of Capgemini Invent France, of the reason for launching the new collaboration.

“To harness the power of this technology responsibly, it is essential to put in place clear and transparent methodologies, and to work with all players in the AI value chain.”

The environmental impact of Gen AI

A recent research paper from Capgemini highlights the significant environmental impacts of Gen AI and agentic AI that operate from operational usage, involving electricity and water consumption and the manufacturing and end-of-life processes of hardware equipment. 

Capgemini points to the large-scale infrastructure development and manufacturing needed to support the growth and adoption of AI, including servers, data centres, telecommunications networks and end-user devices. 

Collectively, these contribute to global environmental impacts such as greenhouse gas (GHG) emissions, resource depletions, water consumption and the growing challenge of electronic waste. 

Large Gen AI models are particularly impactful. Capgemini research suggests that the technology consumes up to 4,600 times more energy than traditional models, and is one of the reasons for the rise in GHG emissions in nearly half the organisations involved in its Developing Sustainable Gen AI research.

By developing a modelling approach that considers areas such as AI usage, changes in the electricity and energy mix, and computing efficiency Capgemini finds that, under a high adoption scenario, AI-related electricity use could potentially increase 24.4 times in the most extreme scenario by 2030. 

Despite these forecasts, positive and tangible action among organisations and leaders is slow. Capgemini surveyed 2,000 senior executives from organisations with annual revenue above US$1bn for its Developing Sustainable Gen AI report, finding that only 12% of businesses measure Gen AI’s footprint. 

Further, 74% say a lack of transparency from Gen AI providers makes measurement challenging, while 42% say they’ve had to relook at their organisation’s sustainability commitments due to Gen AI adoption. Only 27% of executives say they compare Gen AI models on the energy implications before selecting one.

How the observatory address Gen AI sustainability

The newly launched observatory will analyse and mitigate the environmental impacts of AI at all stages of its lifecycle: training, adjustment, inference and end-of-life.

It aims to create an effective methodology and approach for the more sustainable use of AI technologies by bringing together a diverse and multi-stakeholder community of international businesses from academia, business and civil society. 

“We firmly believe that research and action must go hand in hand to drive positive change in our societies,” explains Anne Bouverot, Chairwoman of the Board of Directors and ENS-PSL, a French university that defines and applies scientific and technological research.

The observatory will create a global, open-access database where AI developers and researchers can contribute with data on the environmental performance of their models. This will help drive greater transparency and collaboration between businesses and research groups, which is essential for promoting more sustainable AI practices. 

It will also act as a catalyst for bringing together key industry stakeholders and work to provide strategic analysis and disseminate knowledge. 

“This Observatory is the bridge between the two and is part of our commitment to provide informed solutions to public and private institutions on artificial intelligence and its impact,” explains Marc Mézard, Chairman of the AI and Society Institute.

Capgemini offers its own roadmap to more responsible Gen AI for sustainable business value. As priorities, it says organisations should identify the right technology for specific business needs and assess and mitigate environmental impact at this stage and during application. 

Businesses should also develop the right data and technology foundations and develop and set the appropriate governance for sustainable Gen AI now and into the future.

Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. 

Discover all our upcoming events and secure your tickets today.

AI Magazine is a BizClik brand