nothing prevents them to run a gpu locally or on their own infra. I was asking because I wonder what the enterprises that want to both use AI on their workflow

nothing prevents them to run a gpu locally or on their own infra. I was asking because I wonder what the enterprises that want to both use AI on their workflows like LLM's and have air-gap owned 100% data and pipelines are doing rn. Feels like one of the few areas where to compete with big labs to me, might be wrong