that makes sense to me in a conceptual sense
however inference is very profitable and plummeting in cost for a given point on the intelligence curve, and nvidia gpus can serve different models so they are protected post-buildout
however inference is very profitable and plummeting in cost for a given point on the intelligence curve, and nvidia gpus can serve different models so they are protected post-buildout
Is there any actual evidence of that?
Is it? OpenAI and Anthropic are burning cash faster than anyone has ever shoveled cash into a furnace.