Specific to AutoCAD:I'm with the others on this. The answer to this question is close to asking for a string's length, i.e. the answer will differ because of situation. From experience, the "best" addons come from a combination of Macro, Script, Lisp, DotNet and even sometimes ARX. Each has their pluses and minuses, and usually you can use the one's strengths to offset the weaknesses of the other(s).
What I'd say is Lisp is a definite in every instance, if only to make your setup as easy as possible. Nearly all ARX's I've seen use some Lisp at least to integrate them properly into acad, even most DotNet use Lisp in this sense, if not Lisp then at least some macros with registry hacks. Some things are simply too difficult and/or time consuming to do otherwise. On the other hand some things are near impossible to do purely in Lisp (well that is
AutoLisp).
As a general rule, when your addon is doing something simple like running some commands varying only on some user input or values of standard objects - then Lisp is 10 times easier and faster to use as your language (as compared to DotNet, exponentially more for ARX). If you're working on special objects / features added to acad after around 2000, then Lisp (even with ActiveX) tends to become cumbersome (sometimes even impossible) as the library functions haven't been updated to suit these new objects / features, in some rare cases even DotNet is not enough. Most "involved" addons would fall somewhere in between those 2 extremes, but I'd say 90% of them don't need anything more than Lisp, and 60% would be better suited to be written in Lisp.
Concerning the move to cloud based: if such is the case, then DotNet & even ARX might see a huge decline also. Probably in favour of JavaScript, but perhaps ASP.NET might be used.
In general:Note: in this section Lisp refers to the wider group of languages falling under the Lisp family (Common Lisp, Scheme, Clojure, newLisp, QuickLisp, Haskel, Nyquist, etc.) See the wiki on that:
http://en.wikipedia.org/wiki/Lisp_programming_languageAs for "Lisp dying" ... that's an old misconception. It's been said since (at least) the '80s, in relation to many a now "dead" language. And has been proven wrong every time. There's even been a revival of Lisp in recent years, at least as a training language in universities for programming students, though they tend to use full-blooded Lisps instead of the hamstrung AutoLisp dialect (usually using Scheme or some such) - note many a concept has been stripped out of AutoLisp (e.g. Object Orientation, Macros, BackQuote, etc.). The reason I think this happens is because Lisp is an extremely simplistic language while also being an extremely powerful one, few other languages can say the same. Nearly all programming concepts and paradigms are possible (if not originated) in Lisp.
Also, if you're stating that C# is going to take over because it's "newer" ...
... that's just wrong. First C# is based on C/C++, which in turn was based on Fortran - which is in fact older than Lisp (by a year). Nearly every Lisp dialect these days can say the same thing about their heritage, i.e. created somewhere in the 60s to 2000's, but based on something which was first envisioned in the 50s. Second, simply because something is "new" doesn't mean it's going to be with us for a longer time than something older - e.g. VBA is already in its death throws, even though it was much younger than AutoLisp. The true longevity of a language is a factor of 2 concepts: (1) who and how many programmers use it; and (2) who controls it. I think both Lisp and C/C++ would outlast C# by decades (if not longer) because they have something over C#: they're not owned & controlled by any one single organization (and that being the same org which decided to slowly put VBA out of their misery).
