Hard to find super specific numbers, but 4j per token seems to be a common and reasonable eating, which would mean most chats are a few Wh or less. This paper goes into it a bit more - arxiv.org/pdf/2310.03003

Kind of odd to see the “CUDA is dead because ptx can be written to be much more efficient” takes, but it the DeepSeek paper is an interesting case study in how constraints breed creativity.