Reminds me somewhat of the Minsk- series of indigeneously developed Soviet mainframes -- stack architecture, RISC-like ISA, far ahead of its time. Due to pressure from the higher ups -- or as my parents bluntly put it, "комуняки" (the commies) -- it was abandoned in favour of reverse engineering and cloning the IBM 360 (an enormous feat of engineering) an inferior architecture (my dad called this a setback for Computer Science[1] -- not just in USSR, but in general).

Read more here: http://www.computer-museum.ru/english/minsk0.htm

I guess "no one got fired for choosing IBM" held true not only under capitalism, but under communism as well.

(In a "small world coincidence", I happened to work with the son of the designer of the Minsk-32 -- the most advanced machine in the series -- at Yahoo. Belarus was one Soviet Union's Silicon Valleys, and incidentally the Belarusian State University Institute's Radio Technical Institute still has an excellent EE/CS program (hint: I think you'll have a lot less competition recruiting there than at Stanford!)

[1] Personally I'm a bit skeptical of stack architectures despite their huge popularity (JVM, CLR, Forth, etc...):I understand the cachet, but (to also quote my dad...) "programs can play chess very well", i.e., register windowing is one of the most well understood and mature algorithms, let the compiler handle that job. LLVM's use of a register-based IR is a good step in that direction.