[jkhdigital]

Still doesn’t quite justify homomorphic encryption if the computations involved are mundane (like tax calculations). The true use case is when the computation itself is proprietary, so that users don’t want to reveal their data but the provider also doesn’t want to divulge their algorithm. This is why ML models or finely tuned search engines are the examples most commonly cited, but of course those are also far outside the scale that could be achieved.