A big part of that mistake was believing Motorola and IBM could be partners. Those two plus Apple getting together...you couldn't find three dissimilar tech companies with entirely different ideas or priorities. They tried, with Somerset, but the writing was on the wall very quickly that things were likely going to go the way they did. Scully had no choice but to go down that path though given the choice made before he was there to hitch Apple's wagon to Motorola in the first place.
Mistake or not the move to PPC was a logical one and the decision by Jobs to stick with it jumping to the G3 was also kind of a no-brainer at the time. There was no other move to make that offered a transition to a real, modern OS. He already had to "cut bait" and cancel the failed internal effort to update OS9 to something worth keeping around, sent the Newton team packing, there was simply no way to move to Intel when he took hold of the company again until some years later when it was the only decision that made any sense at all. Once users were weened off OS9 they had options. Not before.
That took as long as it did largely because big software developers are lazy (ie. Adobe being horribly, horribly lazy).
Motorola was/is an engineering-centric company that doesn't understand software, or anything much beyond the component level use of their tech and whose bread-and-butter was non-PC implementations of their technology. Apple might have been the most high profile public PPC customer but they were just a fraction of the PPC market. They were the tail trying to wag the dog. That and, though they're not unique in this, they (Motorola) were and are a company run by Lumberghs. Freakin' idiots. And so they will continue to spin off parts of their company that have any value until Motorola will simply cease to exist. Because they're idiots.
IBM could also not give a crap about much of anything but embedded applications which is why they didn't care about Altivec regardless of the fact that Apple's products were more than a little dependent here. Embedded applications didn't need that kind of floating point acceleration. The rub was IBM had the better manufacturing technology and could deliver a better version than Motorola, they just didn't care.
The switch to Intel was something he had in his back pocket the entire time because NeXT had already been running on "white hardware" for years and OpenStep was already powering a majority of the behind-the-scenes enterprise level web technology. Apple was coy about this for years for the sake of their tenuous alignment until it was simply idiotic for them to stick with it given how utterly disappointing the G5 was and no indication that they could ever get back to being competitive in markets they had been dominating that were starting to slip.
