I learned it because of group projects where we could pick the implementation language and that's what our group wanted.
At that time, Simon Peyton Jones was employed by Microsoft and, at least as I perceived it, was allowed to work on ghc full time because of the goodwill it generated amongst programmers for Microsoft to be the patron of Haskell.
Also, Microsoft is quite open in its support for more heterogenous programming than typical Unix environment, with significant use of "niche" programming languages in various places (including things as atypical as putting custom Prolog implementation into Windows network configuration tools, just to run a prolog program that handled said configuration)
With the advent of Learn You a Haskell, Real World Haskell, Haskell Book...etc, I really feel it is a lot more popular now, but I don't have your experience to draw upon so maybe it is just cognitive dissonance on my part.
I doubt Paul Graham was into the theoretical programming scene in that era. He might have been aware of it, but I doubt it was as production worthy as Common Lisp and thus probably not worthy of his time.