How useful is hardware knowledge in AI s-risks research?

Recently I'm thinking on major in Electrical engineering(EE) vs CS, I think hardware knowledge is less relevant to AI s-risks research, however, it seems EE engineers would have more expected earnings than CS, which is better for earning to give(since hardware would be replaced later by AI than software).

80000 hours has put " AI hardware engineer" as a  career path[1], but there isn't much other discussion about it now.

In your opinion, how useful is hardware knowledge in AI s-risks research? It's probably less useful than CS courses, but how less useful is EE-specific courses(such as eletcromagnism, circuits), compared to CS-specific courses?(here I don't mean the fundamental maths, more like CS-specific courses, for example: Computer Network, operational systems)

Thanks for your answering a lot.

[1]https://80000hours.org/career-reviews/become-an-expert-in-ai-hardware/