Indeed. Cheaper to keep COBOL folks on staff and the mainframe running than to rewrite it (millions vs tens or hundreds of millions of dollars in costs).
Honestly, that is not true. As someone who had to interface with these systems, the cumulative cost of keeping them running is far greater. There are number of layers which run of top of them to support modern age use cases. The problem is no CEO/CTO at big banks who wants to keep their job is going to pull the trigger and transform it, as most likely that project will run over budget and is risky.
You also need to pay to keep the new system running. That’s the real ‘hidden’ cost that you need to be aware of. COBAL may not be sexy, but it’s stable in ways most new systems can’t hope to match.
The experience in the UK of the TSB porting to some Java system written by Santander is a good lesson of why this is risky
You seem to be implying risk is not a cost.
True. And if you decide to re-write the software then you will be eternally re-writing that software to keep up with whatever is trendy.