#pandas has two internal ways to store strings: NumPy and PyArrow (faster).
pandas 3.0 will change the default and strings will use PyArrow when for example calling read_csv.
You can get this change now in pandas 2.1 with:
pandas.options.future.infer_string = True