Story in this morning's paper about how several big-ticket weapon systems are greatly over-running initial cost estimates. I have seen this issue up close, and it is definitely a problem. More correctly, it is a series of problems.
First, there's the requirements definition process. Services (Army, Navy, Air Force, Marines) buy things...it really is one of their chief functions. Service requirements definition processes aren't too far off from what many have gone through when they try to build a house....lots of changes to the plan, and each of those changes costs more money. Because it takes so long for many of these systems to mature, changes in technology greatly outstrip the capacity to deliver the capability. New toy comes along and folks want it included.
Next there's the contracting process itself. More oversight must be placed on the front end, ensuring that the cost data developers put forward is actually attainable. There's an uneasy relationship between the service program managers who really, really want to field the capability, and the contractors who really, really want to build. Both have a vested interest in keeping cost figures as low as possible, at least until you've reached a point of no return.
Finally, there's a leadership problem. Very few people ever get fired because their program had cost over-runs. Oh sure, every now and then the egregious offenders get sacked. But for every program manager whose firing makes it into the press, there are dozens more who have lorded over programs with cost over-runs that just don't make it to the eye-popping level.
This is fixable, but it will take time and leadership.