The main reasons are probably for security and redundancies, a lot of the stories you hear of hackers who manage to infiltrate some site are over rated because the critical data that once used to be central are now spread all over and each site has it's own security firewall so even if any hacker whether inside or outside manages break into 1 site and got hold of the data, it's just a meaningless bunch of numbers and they won't know what it means or what the value of the information really is.

redundancies for business continuity, after sept 11 2001, a lot of major corporations started to change their DR strategies. Hence you end up with many many boxes everywhere.

Btw managing 100+ servers is not easy job, is very good to wish but believe me it's tough to handle it all by yourself. Just imagine you have to install latest sun recommended patch clusters across the landscape, you start friday night you probably only be done monday morning. Or example DBA team wants to upgrade database and tells you to upgrade the OS, security team wants you to harden the boxes and apps team have some change request your work will never get done because every weekend you'll be burning midnight oil and you'll burnout under 1 year.

The optimum number of servers per engineer is usually between 25-40 100 is too much for any a single engineer to handle alone, a team of engineers maybe.

my 2 cts.