Too late....
Ten (hm.... twelve?) years ago this would have been a vital piece of information. In our company we did use IBM "mainframes" (yeah, no longer the bakkelit and steel kind, sadly enough) until then. For political reasons we were pushed to use the Siemens system (ok, make that more like.... 15 years ago), since it was a more local company. Migration started. Then they sold it to Fujitsu. Then Fujitsu said it won't continue development, and couldn't we all stop using those systems. Then they started charging higher prices for the systems. They are usually leased. Effort is underway to migrate everything to x86 based systems, but it is a large application landscape, and apparently the very well thought out assembler code (and the COBOL code) are really very finely tuned, and the modern systems struggle to achieve the same speed in their calculations. Might have something to do with the more "modern" (don't make me laught) programming languages, and the lost appreciation for cpu power and memory being limited ressources. Guess what, when you need to do a lot of these calculations concurrently these things do matter.
With IBM's current offer we could have (maybe...) transferred a lot of the legacy code to x86 systems quite a bit faster. The need to rewrite stuff for the 21st century (so that not only greybeards can maintain it) still would be there, but we could get rid of that "mainframe" system rather sooner than later. Cost wise it would be appreciated.
Cost-wise it is nonsense to use Oracle for everything new. Basically it negates the cost savings of the migration.... *head@desk*