Maintaining servers is often costly. Electricity is a point and complexity is another: various technologies, various network connections, etc. The use a few years ago was to have 1 server per business applications.
The flaw of IT services is often to just let things how they are until something bad happens. But losing money day after day is a bad thing, and you can do better with a strong Supervision process. Supervision of servers must include graphs of intensive values: number of connected users, CPU usage, memory usage, inbound network flows, etc. With these graphs, you can identify:
- Low use deprecated servers and effectively unused servers (happens, sometimes), which you can decide to just stop.
- Low charge but important servers, which you can virtualize. You'll then reduce hardware costs and decrease complexity through homogeneity.