I've always had this idea in my mind that the 'legendary failure rate of IT projects' is best explained in terms of an exponential increase in points of failure

I've always had this idea in my mind that the 'legendary failure rate of IT projects' is best explained in terms of an exponential increase in points of failure (or even just points of activity) when you deal with computer systems. Are there any famous papers that flesh out this idea? Anyone?