My PI once said that the reason for why direct-PhD is so prevalent in the US is some decision made by NSF (or another grant-awarding agency) somewhere in the 60s or 70s that affected the way professors received grant money to fund graduate students. If a professor asked for, say, $25k to fund a masters student, the prof wouldn't get anywhere close to $25k, if the prof got anything at all to pay for a masters student from the grant, whereas if the prof asked for the same $25k to fund a PhD student, the prof will get a sum much closer to the amount originally requested, although not always the full amount. The NSF claimed back then that PhD students were more productive than MS students.
Perhaps that is correct, perhaps it is not.
It is undeniable that BSc+MSc+PhD is better from a student perspective, but the time to a PhD from BSc is usually longer in MSc+PhD systems than in direct-PhD ones, assuming fast-tracking is not in use.