Capgemini: Data Centre Emissions Surge from Gen AI Demand

Global consulting and technology services provider Capgemini has reported that corporate data centre emissions are increasing due to the adoption of generative artificial intelligence (Gen AI).
This is inevitably creating challenges for enterprise sustainability targets.
The firm's research reveals that 48% of executives have observed increases in greenhouse gas emissions following their implementation of Gen AI systems, which use machine learning to generate new content from existing data.
The findings emerge as organisations worldwide grapple with balancing technological advancement against environmental responsibilities, particularly as regulatory pressures mount for carbon reduction.
Data centre power consumption to reach 130GWh by 2030
The computational requirements of Gen AI systems are creating pressure on data centre infrastructure. Industry forecasts indicate that AI development will drive annual data centre power consumption growth of 15 to 20%, reaching between 100GWh to 130 GWh by 2030 - equivalent to powering two-thirds of US households.
The manufacturing of graphics processing units (GPUs), specialised computer chips designed to handle AI workloads, requires rare earth metals. The mining process for these materials generates additional emissions whilst depleting natural resources.
Current projections indicate Gen AI could generate between 1.2 to five million tonnes of electronic waste by 2030, raising concerns about disposal and recycling infrastructure capacity.
The operational phase of these models, known as inferencing - where the AI system processes new data to generate responses - demands equal or greater energy resources than the initial training phase.
The environmental impact of training large language models (LLMs), which power Gen AI applications, continues to expand. Training GPT-3, an LLM with 175 billion parameters, consumes electricity equivalent to 130 US households annually. Its successor, GPT-4, requires power comparable to 5,000 US homes per year.
Water consumption presents an additional environmental challenge, with each set of 20-50 queries to an LLM requiring approximately 500ml of water for cooling systems. This water usage occurs primarily in data centre cooling systems, which are essential for maintaining optimal operating temperatures for AI hardware.
Despite McKinsey research indicating 65% of organisations use Gen AI, only 12% of executives measure their AI carbon footprint, while 27% compare energy consumption between AI models.
Performance, scalability and cost remain the primary decision-making criteria, with only 20% of organisations considering environmental footprint among their top five factors when selecting or building AI models.
“Organisations need to make the footprint of Gen AI visible within their business analysis,” Steven Webb, UK Chief Technology and Innovation Officer at Capgemini.
“Its vital businesses fully track the impact of Gen AI as this enables strategies to measure and mitigate appropriately, as well as unlocking opportunities to implement sustainable practices throughout the Gen AI lifecycle.”
Gen AI use will need to be optimised further
The research from Capgemini indicates that measurement frameworks and industry standards for assessing AI environmental impact remain underdeveloped, creating challenges for organisations attempting to benchmark their systems' efficiency.
In Germany, Compliance Solutions has developed an environmental, social and governance AI system that automates research, evaluation and reporting processes, demonstrating potential applications for sustainability-focused AI implementations.
Steven adds: “Gen AI is already being used to support sustainability in enterprise. With its ability to transform the workforce by automating tasks and complex processes, Gen AI in the form of AI agents can play a pivotal role in optimising use of resources and improving efficiency at all levels, which is key to accelerating sustainability.”
The report emphasises the need for standardised measurement and reporting frameworks as AI deployment increases across industries. This standardisation would enable organisations to make informed decisions about AI implementation while considering environmental impact.
Cyril Garcia, Capgemini's Head of Global Sustainability Services and Corporate Responsibility and Group Executive Board Member, says: “If we want Gen AI to be a force for sustainable business value, there needs to be a market discussion around data collaboration, drawing up industry-wide standards around how we account for the environmental footprint of AI, so business leaders are equipped to make more informed, responsible business decisions and mitigate these impacts.”
Explore the latest edition of Data Centre Magazine and be part of the conversation at our global conference series, Tech & AI LIVE and Data Centre LIVE.
Discover all our upcoming events and secure your tickets today.
Data Centre Magazine is a BizClik brand