The emergence of AI has both raised concerns that AI-fuelled data centre growth might fuel climate change and also raised expectations that AI applications in the energy sector could help reduce emissions by unlocking new optimisations and efficiencies. As over 100 countries – and the European Union – have targets to reach net zero emissions between 2030 and 2070, it is pertinent to explore what AI’s impact on emissions could potentially be. 

Global fuel combustion CO₂ emissions are estimated to reach 35 000 million tonnes (Mt) in 2024. Data centres account for around 180 Mt of indirect CO₂ emissions today from the consumption of electricity, not including any emissions from backup power generation. This includes all workloads by data centres, of which AI is a subset. Data centres therefore account for a small share of emissions: 0.5% of combustion emissions today. Indirect emissions from data centres grow by almost 80% over the course of the decade, rising to 1% in the Base Case. They grow 2.5 times to reach 1.4% of combustion emissions in the Lift-Off Case.

While the share of data centres in aggregate emissions may appear small, data centres are among the few sectors – along with road transport and aviation – that see an increase in their direct and indirect emissions to 2030. In the Lift-Off Case, data centres see the largest emissions growth among all sectors. 

AI applications in the energy sector are being used for a wide range of optimisations, some of which lead to emissions reductions, whether directly through reduced energy needs or otherwise:

• Methane emissions reductions in oil and gas operations – a large source of this sector’s methane emissions come from leaks; AI can facilitate detection so that repairs can happen sooner, for example through better identification using satellite monitoring systems.
• Power sector emissions reductions by improving efficiencies at fossil fuel-powered plants; for example, by ensuring process conditions within a natural gas-powered plant are closer to those for optimal efficiency.
• Industry emissions reductions by optimising manufacturing processes for their energy needs, therefore lowering related emissions; for example, improving the fuel mix for cement production can improve energy efficiency by more than 2%.
• Transport emissions reductions through more efficient vehicle operations and utilisation; for example, improved route choice or driving characteristics lead to efficiency gains of 5-10% and hence reduce emissions.
• Buildings emissions reductions by optimising energy consumption in buildings equipped with management systems; for example, an optimised heating, ventilation and air conditioning control can save around 10% in energy consumption.

The adoption of existing AI applications in end-use sectors could lead to 1 400 Mt of CO₂ emissions reductions in 2035 in the Widespread Adoption Case. This does not include any breakthrough discoveries that may emerge thanks to AI in the next decade. These potential emissions reductions, if realised, would be three times larger than the total data centre emissions in the Lift-off Case, and four times larger than those in the Base Case. 

It is vital to note that there is currently no momentum that could ensure the widespread adoption of these AI applications. Therefore, their aggregate impact, even in 2035, could be marginal if the necessary enabling conditions are not created. Barriers include constraints on access to data, the absence of digital infrastructure and skills, regulatory and security restrictions, and social or cultural obstacles. They could be negated by rebound effects, such as those enabled by modal shifts away from public transport towards autonomous cars. The net impact of AI on emissions – and therefore climate change – will depend on how AI applications are rolled out, what incentives and business cases arise, and how regulatory frameworks respond to the evolving AI landscape.