First of all, Haskell will not die as will not Standard ML and Ocaml or mathematics, because it's based on eternal mathematical constructs and concepts (up to isomorphism :-)

The question is whether it will be utilized to the full potential and what is needed for that. It is of course easy to work in some framework and just put well-defined pieces to intuitive places to create software. It's harder to create application architectures, which allow to put well-defined pieces to intuitive places.

The first of these two will be the domain of AI quite soon, so developers will be freed to do the second and be on the bleeding edge.

As already mentioned above, Haskell (most of the time) is built on mathematical structures, which become more an more standard even across mathematics, so it is inevitable that something Haskell-like will be a big thing in the near future. It does not really mean that Haskell syntax will survive as such, but in my opinion some kind of "Python" can come out of it. Like Elixir for Erlang.

While I am not using Haskell in production, I studied it and it helps in programming elsewhere.