My bet is that Anthropic is a big costumer of theirs (they implied 'AI' companies are very interested) and they want their supply chain to be 'secured' with their own product.

You can listen to their podcast on AI. They are interested in it and use it they are not AI pilled.

LLMs come up a good bit on the Oxide and Friends podcast. I'd recommend giving it a listen. They have very reasonable takes on LLM usage.

I think the idea of being all in on LLMs or being 100% anti-LLM are both positions that miss opportunity for better development.