How to calculate cost per tokens?
I just wanted to know how much it actually costs for OpenAI and Claude to host the LLM models and how much they are subsidizing, and whether it is actually feasible for them to host models at the prices that they are. I just wanted to know how much it actually costs or what sort of calculations should we do to understand the unit economics of generating tokens. I know that most of what OpenAI and Claude might be doing are some optimization techniques that are not known to the public. But can you help me to understand calculations by explaining it for DeepSeek hosted on some cloud platform like AWS? I just want to understand what sort of calculations I must do. When I started to do it, I got confused about where I should start from. And what things should I consider?