What Ethereum marketing likes to call a “global supercomputer” is slower than a 1950s computer built out of vacuum tubes, and running programs is more expensive than it was in 1955. (Programming ergonomics are also nearly punch card level.)
There are very few applications you’d actually want to execute on a platform like that. And every attempt to improve the design (including the long-promised PoS transition) is going to turn it into a centralized plutocracy.
It’s a fascinating science experiment that doesn’t have much purpose as an application environment. But because people are stubborn and the marketing was successful enough, “blockchain” is to this decade what “object-oriented” was in 1990s: a token gesture towards it will make managers and journalists happy.