Why is vectorised NumPy faster than a Python loop?

An answer that works

The loop executes in the Python interpreter with per-element overhead, while the vectorised operation runs in compiled code over contiguous memory, often using multiple values per instruction. The difference is typically one to two orders of magnitude.