An LLM neurone is a software abstraction. It has nothing to do with GPU logic gates. In fact you dont even need GPUs specifically to run an LLM.

You shouldn’t compare biological neurones to microprocessor logic gates because that’s like comparing a car to a pair of shoes.