The problem is that for many areas of software it is not possible to be effective without at least a couple years of fairly deep experience. That is the learning curve regardless of how smart someone is and it is difficult to justify hiring someone that will not be qualified to fill the job they were nominally hired for for two years. It is not a matter of training per se, it is the amount of time required to develop competency.
For many areas of software development, such as database engine internals, the pool of people that are qualified or trainable within some reasonable number of months is quite small. Therefore, for those areas of software development the shortage is real for all practical purposes. A PhD in computer science does not automagically confer the ability to become a software subject matter expert within a few months.
The problem is that for many areas of software it is not possible to be effective without at least a couple years of fairly deep experience.
This point is key given that industry turnover is such that the first company to hire you has to expect that they are simply training you for the next.
On the other hand, a for such a specialized area a relevant PhD seems almost a necessity. Either that, or an equivalently long time in industry specializing in that area.
I doubt that this is the norm though. I get the impression that most of these hiring managers who don't value experience are hiring for pretty uninteresting work and simply want inexpensive labour. The op didn't state what area of work he's in, but if a junior is the ideal candidate, it's very likely not very interesting or difficult work.
edit: after a bit of digging, it turns out that the OP works on trading systems in Canada. The ironic thing is that 125k is quite a bit above a junior position in most of Canada. So it's not clear what is going on here. When hiring PhDs, he's competing with the US companies and might be a bit shy of what they pay. On the other hand, if he's paying a junior 125k, he's paying well above the local average.
For many areas of software development, such as database engine internals, the pool of people that are qualified or trainable within some reasonable number of months is quite small. Therefore, for those areas of software development the shortage is real for all practical purposes. A PhD in computer science does not automagically confer the ability to become a software subject matter expert within a few months.