Actually, supporting legacy systems is a VERY lucrative job.
Businesses most certainly grow to large to even begin consider updating systems. Do you have any idea how many mainframes from, say, the 90's, are still used in some of the largest institutions in this country? Why? Because so much of their business flows through the systems it is disastrous, if not fatal, if there were even a single day of disruption.
What is the one constant with new systems? New bugs to be worked out. A good friend of mine does software consulting for one institution for which about every single day $1+billion dollars flow through the software.
It's all fine and dandy to scream upgrade at every opportunity, but that just means you're not involved in anything mission-critical.
Lol. Mainframes? They're about the only exception to this rule.
Software vendors completely stop supporting outdated operating systems. Linux, Windows, Mac... they all eventually cut ties and say it's time to move on.
I won't once suggest (and even eluded to it in my prior messages) that one should upgrade at EVERY opportunity, but to stick with software that is over a decade old is simply lazy. I've seen vendors hold companies back to IE8 because their app 'won't work with anything else.'
And while supporting legacy systems such as a mainframe can be lucrative, supporting a legacy Exchange 2003 environment is beyond pointless when there are far better systems that are actively being improved on the market.
Suffice it to say, I work for a service provider and we have a customer who is using Exchange 2003 by their choice. Support is difficult, cumbersome, and data loss is general possible and nearly likely as supporting that system becomes difficult when MS won't take a phone call and the backup platforms have phased out support for it long ago.
You can tell me all you want about a consistent software platform being the answer to longevity, but it truly is not. Linux LTS releases are typically good for 5 years. MS supported XP for roughly 13.
Lastly, running a multi-million dollar business on 15 year old gear that very few engineers know how to support is simply a point of accepting too much risk. At some point, that system is going to go down, and the odds of bringing it up can become very slim.