Right. Even if they do keep the spend up, we may still soon end up at a point where a cheap open model is good enough. It could be hosted locally or provided through any of the cloud providers without any revenue going to OpenAI or Anthropic.
They have some stickiness to the tooling people are used to, but those can be overcome when there's a significant price difference for the same results.
I don’t think the more is ever going to be the actual LLM’s, but the harnesses and support around them. If Anthropic can own the model and the harness setup, I could see there being some efficiency gains there that would be worth something.
