Is there an interest in hosting GHC on the JVM (besides my own).

There's interest but my understanding is that the GHC backend architecture is not at all friendly to work with.  That said, I hear in the next release (I think 6.12, not the 6.10 that's in beta) will have a redesigned backend architecture that is supposed to be much easier to work with, which will make it easier to provide native code generators for many currently poorly-served platforms --- including the JVM if anyone is so inclined.

I don’t think it’s so hard to translate GHC’s Core language to the JVM.  Apart from the treatment of tail recursion, for which various hacks are available.  Eg Scala manages it.  I don’t think you want the C-- level (which is what the “redesigned back end” is about).

However, there’s more to it than generating bytecode:

http://haskell.org/haskellwiki/GHC:FAQ#Why_isn.27t_GHC_available_for_.NET_or_on_the_JVM.3F

It’d be a good project though!

Simon