(In a previous life I worked at Bloomberg) I don't remember the exact figures but 25m lines of Fortran seems plausible, but it was still the minority of the co

(In a previous life I worked at Bloomberg) I don't remember the exact figures but 25m lines of Fortran seems plausible, but it was still the minority of the code base. The majority at the time was in C and C++, with probably a few million lines of Perl and Javascript as well. Plenty of the Fortran dates from the 80s, rewriting it in a modern language would be a huge project which lots of risks and limited upside. No significant new fortran has been written for a long long time (>10 years). It's also a mistake to think of Bloomberg as a single piece of software, it's closer to an app store, there's are thousands of specialist apps (functions in Bloomberg terminology) and most users only use half-a-dozen, but which half-a-dozen varies significantly between users. This makes it very hard to displace wholesale. There are individual pieces a competitor can go after but identifying a subset of functionality that's enough to convince users to switch is hard. There's also a lot of stickiness which goes beyond pure functionality. Network effect and brand are also key parts (having a Bloomberg Terminal is a status symbol).