Estimating the growth in emissions from AI data centres

New research projects a substantial increase in CO2 emissions stemming from the expansion of AI data centers, primarily driven by the advent of agentic AI. Utilizing advanced life cycle assessment models and industry-promoted scenarios from the IEA and McKinsey, the analysis indicates a dramatic rise in overall emissions, highlighting the significant contribution of embodied carbon in addition to operational emissions.

Why it matters

This research provides a quantified assessment of the environmental impact of AI data center growth, which is critical for strategic planning in technology infrastructure and sustainability. Understanding these emission trajectories is essential for aligning technological advancements with climate goals and assessing the long-term societal and environmental footprint of AI development.

Key insights

• The emergence of 'agentic AI' is a primary driver behind the rapid expansion of data centers.
• State-of-the-art life cycle assessment models were used to estimate both operational and embodied emissions from AI servers.
• Projections based on scenarios by the IEA and McKinsey suggest a dramatic rise in overall CO2 emissions from AI data centers.
• Embodied carbon, representing emissions from manufacturing and infrastructure, constitutes a considerable component of the total emissions from AI data centers.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.14421