Size of the market vs difficulty of the problem.

I wrote a FORTH interpreter in about 3000 lines of assembler when I was in high school. It's not hard to write a LISP and it's entirely routine to write a simple PASCAL-like language in a compiler class.

The simple case I see is that if you completely understood the bitstream format of an FPGA it might not be hard to make something that compiles some kind of code to the FPGA. To compete with current HDL tools you would need something that produces photomasks, the cost of testing the system alone would be astronomical.