Artificial Intelligence: Emissions

Earl Russell Excerpts

Wednesday 23rd April 2025 Lords Chamber

[Lord Vallance of Balham (Lab)]

I thank the noble Lord for that important question, on which I have had discussions with him before. I think it is an important way of measuring it, but the new AI Energy Council is looking at all forms of link between energy and AI. It is worth also noting that the advances in technology mean that the energy consumption is dropping for many of the approaches to new compute and data centres. In fact, some of the chips being designed now may reduce the energy consumption between hundredfold and thousandfold. There is a need to keep an eye on this and to think about what the appropriate way is to measure both the energy consumption and, as the noble Lord rightly says, the broader environmental impact, including water usage.

[Earl Russell (LD)]

My Lords, generative AI is exceptionally energy hungry. The head of the National Grid said in March last year that data-centre electricity demand in the UK will rise sixfold in the next 10 years. While the Climate Change Committee has some headroom in its carbon budgets, what work are the Government doing with it to ensure adequate and accurate calculations can be made of AI’s future energy requirements?

[Lord Vallance of Balham (Lab)]

The current estimate for 2024 was something like 7 terawatt hours of consumption; if we go forward to 2050, that is expected to be something like 62 terawatt hours, but, as I have just mentioned, the advances in technology may change that. There is a lot of uncertainty around the requirements. It is worth noting that, over the same time period, the increase in energy consumption from many other areas, including from electric vehicles, means that the proportion taken up by data centres, even if there is no technology improvement, is probably something like 10% of the total.