Anthropic just started hiring chip engineers, not model researchers

Anthropic just started hiring chip engineers.

Not model researchers. Chip engineers. The company wants to co-design the silicon Claude runs on, not just rent compute from its existing partners. That is a lab deciding the API layer is not enough control over its own product.

I watch this from the other end of the stack. Every tool I use to design and ship work sits on top of somebody else's model. That has not changed today. But it shows where the leverage is moving. Not toward the apps built on these models. Toward whoever owns the layer underneath them.

If your product runs on an AI lab's API, the real risk was never a pricing change or a rate limit. It is the lab deciding your layer is worth owning too.

What happens to your product the day the model you depend on decides to become the whole stack instead of just the brain inside it?