About 6 years ago we adopted Scala for our server side language, and for the past 4 years we have been migrating old code that was CPU/memory hungry to go and getting very good results. All new code, if it's going to be cpu/memory intense, we write it in Go.

For non C/C++ developers, Go is much easier to work with and get performance benefits that before would only be possible by porting to C, or you end up not writing idiomatic Scala, which defeats the purpose of using Scala (imho)

I guess you meant outside of google? Google's logs analysis is often done in "Lingo" which stands for Logs In Go. I assure you that logs analysis is very CPU bound.

http://www.unofficialgoogledatascience.com/2015/12/replacing...

In most cases likely not, but there are some people using Go for CPU-bound stuff. There was a guy at gophercon yesterday who did a pretty cool demo of a raytracer he wrote in Go. He talked about the 15x performance improvement it had over using Javascript. I would definitely not choose to build a raytracer in Javascript, and probably not even in Go either. C, C++, Rust, are probably the most common choices made by people who are building something that is seriously CPU-bound.