Have a look at what Walter Bright [has to say about C++ compilation speed](http://www.drdobbs.com/cpp/c-compilation-speed/228701711).
The biggest problem that C++ faces in general is that it depends on #include files, which in the worst case can make compilation times quadratic as a function of source code size. Nim uses #include files only sparingly (primarily "nimbase.h" and <string.h>, plus whatever is needed by extern procedures). Each Nim source file, of course, is processed only once. The biggest issue with compiling large code bases right now is the lack of incremental compilation; while the C compilation stage is already incremental, the Nim compiler still has to process the entire source code each time. See [this GitHub issue](https://github.com/nim-lang/Nim/issues/7874) for Araq's plans for incremental compilation in Nim.
