Like @jetti, I would (and have) gone with what I already know. In my case, Groovy and Java. There are other languages I'm intrigued by, but I haven't seen enough evidence to convince me that any new language is going to offer such radical improvement in (performance|productivity|stability|whatever) as to justify the huge time sink it would be to switch.
That said, we try to make things API driven as much as possible (I'm loath to use the term "microservices", but I guess it would fit) so we have the option to introduce new languages on a piecemeal basis if/when it seems justified.
Some languages I've played with / looked at (or used in the past) that might make it into the mix someday include C++, Scala, Clojure, R, Python, Prolog and Julia... possibly others as well. But we're not going to start throwing new stuff in just because it's new and trendy. We try to avoid FDD (Fad Driven Development).