The same AI systems positioned as enablers of climate solutions are simultaneously increasing global demand for electricity, water, and critical materials. As a result, the digital infrastructure underpinning the green transition is emerging as a significant environmental load in its own right.
Recent analysis indicates that deploying AI across power, transport, and food systems could reduce global emissions by an estimated 3.2–5.4 billion tonnes of CO₂‑equivalent annually by 2035, primarily via grid optimisation, demand management, logistics improvements, and efficiency gains in production and consumption. AI systems already support more accurate climate modelling, earlier extreme‑weather alerts, precision agriculture, and advanced building management, effectively acting as a control layer for decarbonising infrastructure.
However, large AI models remain highly energy‑intensive to train and operate, and are typically hosted in data centres that depend on carbon‑intensive electricity and substantial water use for cooling. Projections suggest that AI workloads could, by the end of this decade, approach the electricity consumption of tens of millions of households, with associated emissions locked in unless renewable penetration and efficiency improvements accelerate significantly.
The strategic question is no longer whether to prioritise AI or climate, but how to implement green AI as a design and governance standard. This requires explicit sustainability constraints and metrics for model architecture, data‑centre siting, energy sourcing, and product lifecycle, rather than treating environmental impact as a secondary consideration. Organisations that adopt this approach will not only deploy AI to reduce emissions in external systems but also decarbonise their own digital infrastructure.
The trajectory of climate‑related innovation over the next decade will be heavily influenced by whether AI development becomes both computationally advanced and carbon‑constrained, enabling large‑scale emissions reduction without driving a parallel increase in the environmental footprint of digital systems.
Further reading and references:
• LSE Grantham Institute & Systemiq, “Green and intelligent: the role of AI in the climate transition” (npj Climate Action, 2025):
• Nature, “Environmental impact and net-zero pathways for sustainable artificial intelligence servers in the USA”:
• Jisc National Centre for AI, “Artificial intelligence and the environment: looking ahead”:
