Python is suitable for weekend toy projects, not for things that run at scale. Just imagine that if you write something in C or Java you might need 5 or 10 servers, respectively, but you would need 100 servers for the equivalent Python code - that adds up to a lot of money and energy.

Machine learning scientists like Python because they are not programmers - they are focused on building models, and Python is a quick and easy language to get access to their models that run in GPUs.

There is an interesting story about Dropbox, which coded its system in Python. They hired the creator of Python. They eventually realized that it was an unmaintainable mess, and that Python's typelessness was a major contributor to that: https://www.zdnet.com/article/python-programming-language-creator-retires-saying-its-been-an-amazing-ride/

Van Rossum also said on another occasion (https://slashdot.org/story/354868): "I've learned a painful lesson, that for small programs dynamic typing is great. For large programs, you have to have a more disciplined approach. And it helps if the language actually gives you that discipline, rather than telling you, 'Well, you can do whatever you want.'"