One of the most common bits of lore about Apple in the early 2000s is that they were trying really hard to sell a net platform that users weren't exactly enthusiastically hopping over to. Mac OS X gained market share relatively slowly. Reviews were good, it was gaining applications, but Mac OS 9 retained the lion's share of the actual users for a long time. At the time, I was on the fence about it. My 233MHz iMac wasn't fast or powerful enough for it, and when I got my hands on a better system, what I found is that for every nicety of Mac OS X, there was typically something broken, missing, or simply better in Mac OS 9, for the way I was used to working my computer, which had been forged in Mac System Software 7, and Mac OS 8 and 9.
Thinking back on it, it's no wonder that Power Macs weren't exactly selling like hotcakes in the early 2000s. A big part of the lore is that Quark thought the whole "Mac OS X" thing would just blow over. They really did think this and it's well documented that Quark was pretty reluctant to make any kind of swift move to Mac OS X. Perhaps, for this reason, many of Mac OS X' early users were UNIX switchers, and people who didn't rely on "pro" Mac applications (the ones that typically took longest to get ported.)
One of my theories has been that for a while, in an effort to sell more Macs, Apple was focus very heavily on building dual CPU systems. If you were using Mac OS 9 for almost anything, a second CPU didn't really matter. If you were buying a Power Mac and you didn't intend to use it for OS X, it didn't exactly hurt, but it was money you didn't need to spend on hardware you couldn't use. Most of the systems through the middle of the life of the Power Mac G4 focused heavily on dual CPU configurations, offering choices such as a single CPU 466MHz model, a dual 500MHz and a dual 533MHz, with nothing really tailored to people who were going to be using an OS and apps not helped by multiprocessing.
One of the more interesting pieces of conjecture at the time, from John Gruber, is that it wasn't just Quark that was slowing Mac OS X adoption (and the sales of Power Macs) but the fact that so many people who use those apps aren't the ones who buy their own systems and software. Random Graphic Designer X can't necessarily buy a new Power Mac and switch to a new hypothetical version of Xpress, because those decisions come from on high, and in this realm, Power Mac G4s and Xpress and Photoshop licenses are fleet tools.
I think another aspect of it, something that gets talked about from time to time today, but wasn't mentioned (or perhaps known at the time of Gruber's writing) was that Apple was selling more iMacs than Power Macs, and more iBooks than PowerBooks (2002 Annual Report). In fact, it looks like 2002 was the cross-over year. In both 2002 and 2003, Apple sold nearly twice as many iMacs as they did Power Macs. Things flattened out just a little in 2004, and by 2005, Apple was already coalescing unit sales numbers into "desktops" and "portables." 2004's "Power Macintoshes" number also included the Xserve, making it look that much worse.
A lot of that probably has to do with the digital hub. Apple had started selling the iPod the previous year, in October 2001, and at that time, you still needed a Mac to rip, mix, and burn, or put it on your iPod to go. Apple didn't sell its own digital camera, but iPhoto allowed users to organize digital photos and create albums. iMovie did the same for videos, and at the very tail end of the digital hub concept, iWeb allowed people to create their own web sites easily.
Gruber is correct in that Power Macintosh sales had gone down from 2000 to 2002. Everything went down in 2001, and in 2002, iMacs recovered, but Power Macintoshes did not. I suspect this is about Macs starting to gain more mainstream traction as computers that normal people could buy. Consider that in 2001, Power Macintosh base prices had risen to a (then) all-time high of $1699 for the 733MHz Quicksilver Power Macintosh G4. For your $1700, you got 128MB of RAM and a 40 gigabyte hard disk drive. The highest end iMac configuration was the iMac SE 700MHz, which cost $1499 for a configuration with 256MB of RAM and a 60 gigabyte hard disk. It also had a display, and both systems included a CD-RW drive, a modem, some software, and a keyboard and mouse.
G3 and G4 performance was pretty similar at the time in most "everyday" tasks, but a few of the things consumers did on a regular basis such as working with iTunes and iMovie were said to be helped significantly by the G4's enhancements over the G3. The uptick in iMac sales in 2002 isn't hard to explain though. By August 2002, your $1499 got you an iMac G4 at 700MHz with 256MB of RAM, a much better GPU than last year's iMac G3, and a 40 gigabyte hard disk, with an LCD display and the typical Mac accoutrements of the time. The base Power Macintosh from the same time was $1599 for an 800MHz CPU, 256MB of RAM and a 40 gigabyte hard disk. At $1099, the original eMac when it was available was also a really good deal, with the same 700MHz G4, 128MB of RAM and a 40GB hard disk. The 800MHz eMac, introduced in August for $1399, had the same 800MHz CPU as the Power Mac, 256MB of memory and a 60 gig hard disk. The Mirrored Drive Door Power Macs Gruber previewed in August 2002 brought a lot to the low end of the Power Macintosh lineup, namely a dual 867MHz CPU arrangement, 256MB of RAM, a 60 gigabyte hard disk, and L3 cache on the low end system, things that should have been compelling for power users of Mac OS X, especially those coming over from UNIX or expecting to run several applications at once. Whether or not home users perceived any benefit from that is questionable. If you were going to spend more than $1499 on a system, there were better iMacs to be had, including the 800MHz model in a 256/60 configuration with a DVD burner, and an 800MHz model with a 17-inch display and an 80 gig hard disk.
Between professionals not really being the one to choose their own machines, the slow movement of software developers for professional applications, and a changing shift of consumer and professional buyers, it's not very difficult to see why Apple was having a hard time with the Power Macintoshes at the time. The Power Macintosh wasn't a very compelling product for home if you could get an iMac that was nearly as fast and included a display for less money.
In 2003, things looked better, between the entire Power Macintosh G4 family from late 2002 shipping with dual CPUs that had L3 cache, and the availability of faster eMacs and iMacs. Power Macintosh sales should have picked up by then, especially with the release of Quark Xpress 6 in 2003, however I suspect that what we saw was the trend we now know to be true: consumers started buying Macs in droves. At the time, Apple didn't make nearly as much per unit on iMacs as they were on Power Macs, but they were selling almost twice as many units, and almost certainly getting good margins on them. In fact, a lot of people talk about the Mac's "core market" but I think that this has been wrong all along. Apple had the graphic designers, but they weren't the core market. They weren't buying the most systems, and they weren't buying systems most frequently.
The other thing that you can see happening from 2001 to 2003, and has continued happening, is that the iMac has become a totally amazing computer. Today, the iMac is at least as good as the Mac Pro for the vast majority of tasks. Whereas in 2001, you probably needed a Power Macintosh if you were going to do just about anything with any desktop publishing, photo work, database, video, or anything of that nature. iMacs could do all of those tasks, but not as well as a Power Mac could.
In the early 2000s, iMacs not only shipped with less RAM and slower hard disks, but with much slower versions of CPUs. The split between the G3 and the G4 was considered significant for certain types of tasks, primarily things that involved media and math. The split was even more significant if you were using Mac OS X, where Power Macintoshes were the only option for dual processor systems, and where Power Macintoshes have other features that (in the early days of Mac OS X) made life a lot better.
It would be insanely interesting to see what the split of Apple's sales is like today. Most "normal" tasks simply don't need today's Mac Pro, and for various reasons (including the fact that they've been letting it stagnate with old hardware for uncomfortably long periods of time for the past five or so years) I imagine pros still using Macs are migrating toward the iMacs and MacBook Pros. Even on the Windows side of things where Apple's apathy doesn't impact pro users, there's all manner of talk about how fast and amazing 15-watt CPUs are, and that Microsoft uses Surface Pros to design the next Surface products.
I don't think the true, high end workstation is dying, and I continue to believe that the Mac Pro is one, and that Apple's position building the only "UNIX workstation" you can buy today is pretty unique, but also perhaps less important than it was a few years ago, because most of that market can get what they need either out of Solaris or Linux, or even Mac OS X running on much more modest hardware.
Tim Cook has already said Apple thinks the iPad Pro is the clearest possible expression of their opinions about the future direction of personal computing. Whether that means they think that in the future keyboards are optional for everybody, or just that they see the majority of home and mobile usage moving in that direction has yet to be seen, especially given that the prodigal iPad is not even available for pre-order yet.
So, there's a lot of reasons for a slump in the sales of Apple's professional computers. There's a lot of folklore surrounding it in the early 2000s and most of that doesn't even account for what was happening in the real world, just the technical merits of each system. Considering economic factors and the growing importance of Internet access probably makes iMacs and iBooks look that much better.