Dr. Patricia Scanlon:

I will answer and then pass to Mr. Blanchfield because he is a good person to answer that. I will not go back through my past, but when I was an engineer 20 years ago, we spent a lot of time concentrating on making algorithms pre-AI - it was machine learning, actually - smaller and smaller in order to get them to fit in the funnels. They started using less energy and smaller chip design. That innovation was the strategy to be able to get it to work on flights. It is very hard to tell what will happen because in a lot of companies the strategy is to be able to put a lot of AI here.

A lot of the inference that consumes huge amounts of energies in the data centres can start being local. That is a huge area of research we all need to embrace. Sometimes people talk about the energy usage as if it is scaling exponentially and will not stop. I do not necessarily agree with that because we will become more innovative in how we do this and deliver it. These chips will get smaller, faster and better.