If one of those companies decided to drop decent money on turning SciPy+Jupyter (or Julia) into a proper high-performance CAS, Wolfram would be in huge -- maybe existential -- trouble.
mathematica is already in existential trouble. yes some theoretical physicists supposedly use it (i used it undergrad to do all of the annoying integrals in e&m) but that's such a small market it's basically irrelevant. i can almost guarantee there is no production code running that uses mathematica in some way.
