Right, but when you're at that point in performance considerations you already have a team of specialists working on multiple angles in performance.

And precisely, for ML code all python libraries run extremely optimized natively compiled code. The language overhead is a minimal consideration. And for business domain code language performance is rarely the limiting factor.

In some companies that team size is a very tiny number of 1.

If your team size is 1 then you're not doing yourself any favor thinking about performance beyond basic usability when dev productivity is a far higher priority.