

Except, now, they represent a higher opex cost than even payroll.
That’s usually because the on-prem equiment was hidden somewhere in capex, not opex.
There are a lot of cases I’ve been involved with where AWS comes out way cheaper than an on-prem solution, once you take all the costs into consideration. But there are plenty of other cases where it has turned out to be more costly, especially if some bonehead attemted a “like-for-like” migration with no effort to minimize AWS service costs. Take every on-prem VM and stand up a corresponding EC2? Replicate the same rat’s next of connectivity that’s in the legacy system? That’ll probably cost you.
As long as I can choose the person, I’d do it. Win/win.