Comment by diggan

> the only way to do that is to replace them with AI

I guess money-wise it kind of makes sense when you're outsourcing the LLM inference. But for companies like Microsoft, where they aren't outsourcing it, and have to actually pay the cost of hosting the infrastructure, I wonder if the calculation still make sense. Since they're doing this huge push, I guess someone somewhere said it does make sense, but looking at the infrastructure OpenAI and others are having to build (like Stargate or whatever it's called), I wonder how realistic it is.