I'm pretty sure you can build a startup around almost any mature, real-world tech stack... so, basically, exclude "esoteric" programming languages like Brainfuck, Intercal, Whitespace, etc. Otherwise, I think you could - if you chose to - build a startup around RPG/400 on an IBM iSeries, Fortran, COBOL on MVS on an IBM zSeries, Ada, .Net, Prolog, Pascal, Perl or whatever floats your boat. It's a question of the tradeoffs and the economics of each decision. Can you hire programmers? What are the licensing costs? What about the "ilities," like scalability, interoperability, reliability, etc?
Maybe for some startups the right decision is .Net, or RPG, or Forth or Rexx or what-have-you. But I'm guessing for most startups, there are better choices. I'm also not sure that - even on a per startup basis - there is one ultimately right choice. More likely there are a handful of really good choices, a pile of so-so choices, and a long-tail of fairly ridiculous choices.
Maybe for some startups the right decision is .Net, or RPG, or Forth or Rexx or what-have-you. But I'm guessing for most startups, there are better choices. I'm also not sure that - even on a per startup basis - there is one ultimately right choice. More likely there are a handful of really good choices, a pile of so-so choices, and a long-tail of fairly ridiculous choices.