The new AMD chips in the Framework laptops would be a good candidate and I think you can get 96GB RAM in them. Also if the LLM software is idle (like llama.cpp or ollama) there is negligible extra power consumption.