Warning: This is a bit long
As someone, Paul Graham I think it was, pointed out, selling Operating Systems is a losing business plan, because anything you can come up with, someone else will replicate and release for free. Any future features you announce will be made faster, better, and again, free. That's simply the nature of the business. Everyone, even Microsoft, realizes this, though for some, that realization is from the proprietary side, and from the open-source side for others.
What is potentially profitable is hardware. Hardware will ALWAYS be profitable, if you have the right business and distribution model. Only Dell and Apple are successful at this. What does that make them? Competitors. Fortunately, due to the differences in their business plans (Toyota* vs. BMW) they are not direct competitors.
*some may wonder about the comparison of Toyota to Dell; I chose Toyota, rather than, say, Honda, Kia, Hyundai, or Daewoo because of their amazing assembly capabilities and efficient busness plans, not based on a quality comparison. Toyota makes great cars. Dell does not make great computers.
The one thing I've left out of this whole thing is Applications. Now THERE'S a place in the software world where real innovation can happen for profit. It's much harder to replicate a killer app than it is to replicate an OS. MS Office is actually somewhat decent in its latest Windows incarnation, and great on the Mac (admittedly, they had a few missteps... Word 6, for example). OpenOffice.org simply isn't everything Office is. There, I've said it, and just made myself a target for every OSS zealot out there. Same thing with PhotoShop vs. the GIMP. GIMP is great for basic image manipulation, but for CMYK, or RAW control, batching, and a standard interface, there simply is no replacement for Photoshop. As a result, it's incredibly popular and incredibly profitable. (I'm leaving out applications integrated into the OS, such as IE, especially since Mozilla is practically a given on Linux platforms)
So where does that leave Apple re: MS and re: Dell? Start by MS re: Windows and Office? MS is trying its darnedest to sell an OS, which in the long term cannot possibly pay off. Still, right now, it's a nice stream to milk for all it's worth, and so they will fight to keep Windows alive as long as they can (see Longhorn). That said, Windows exists to sell Office, which is still much more profitable for them (backwards compatability with all .exe's is millions of man-hours of work, vs. a few hundred for all previous .doc, .xls, .ppt, etc). If Windows' marketshare ever crashed, you can bet your hiney that MS would release a Linux version of Office, to keep their most profitable line of software.
So where's Apple in this? Apple, instead of selling OS to sell software, as Microsoft does, sells OS to sell hardware and software. A perfectly valid approach, barring the Linux problem. And make no mistake, Linux, or another OSS OS is eventually going to be on EVERY desktop. It's just a question of how much time and misery the computing world goes through with TCPA/NGSCB first. But Apple, one step ahead of everyone as usual, cuts that off by opening the core OS, Darwin, and, more importantly CoreFoundation. Which can, and does, run on just about every piece of hardware out there, as long as it has a driver for it. By doing this, Apple effectively created a logical partition in between OS technologies and Application technologies
inside of what we typically think of as the OS iteslf! This sets Apple's Darwin up to be the next Linux, a great position to be in under any circumstances. Then, you can pay $129 for the Applications and application technologies that run on top of Darwin: Cocoa, Carbon, QuickTime, iChat, WebObjects, Quartz, Aqua, Finder, Spotlight, Dashboard, iLife, Sherlock, Mail... the Apple Crown Jewels (well, maybe not the Finder

). But, as with any great software package, there's system requirements...
the most important of which is "An Apple Macintosh computer".
They are using the Crown Jewels to sell the premium-priced hardware. Now THIS is an absolutely rock-solid, future-proof business plan (provided they can get the best hardware to run those jewels, hence the Intel switch**), one that leaves absolutely no room for Michael Dell. The most Apple would likely ever do is another rebranding gig if they ever need more manufacturing and distribution capacity, or someone can offer extremely sweet licensing terms, a la HP, whose computers have iTunes installed by default.
**For the record, I think the Intel switch makes business sense for reasons outlined above, but is still a terrible, terrible thing to do architecture-wise, and is going to be much more of a pain for users and developers unless they are willing to speed up Rosetta (currently translates to 800 MHz G3, not coincidentally the fastest G3 desktop computer Apple ever shipped, for technical reasons) and give Classic support a higher priority than "very low". If Apple survives the to-be-painful transition period, and I wouldn't completely confident they will, given how their marketshare decreases with every transition (to say nothing of the apparently aborted transition to 64-bitness), they will thrive in the long term. Apple does, however, have one last ace up it's sleeve, which is probably the one for the next two years: free (as in beer) Cocoa, QTKit, EOF, WebObjects, and WebKit.
First, a bit of background: Cocoa was the Hope Diamond of NeXT, even before it was called Cocoa (I think that's an Apple term actually). A set of Objective-C frameworks written to be architecture independent, with a ton of built-in functionality and an incredible interface builder (called Interface Builder, of course). Way back when, NeXT charged for the use of Cocoa (well, OpenStep anyway), and it's web-oriented cousins EnterpriseObjects and WebObjects. NeXT charged A LOT. When Apple bought NeXT, Apple charged A LOT for the developer version of WebObjects ($50,000 plus deployment license costs), and hid away Cocoa until OS X. With OS X comes the return of Cocoa, now free for anyone to use if they installed the free Dev tools. WO was reduced to $699. Jump forward to now, with Xcode 2.1 and WO 5.3, all of that is free with the free developer tools! Anyone with a Mac can potentially become a Mac developer at no extra charge, whether their target is the desktop or the web! ANYONE! This is HUGE. MS only started doing this recently with .NET, but the IDE and compiler is still $600 or so to a "normal" (not corporate or education, which actually excludes quite a lot) customer. This means that the potential percent of Mac developers is much higher (100% of those running OS X; 16% and growing daily for WO, though Apple did give away rotating trial licenses for WO 5.2 under Panther and Jaguar, so really more like 84%) than the potential percent of Windows developers. This, I think, is what will get Apple through the next two years: new programmers, seeing just how easy it is to get started developing on a Mac, will start developing, especially with the introduction of Core Data and QTKit which make it so easy to do great things. The number and quality of new Apps will make Apple and the Mac platform survive through the next two tough years to the bright future ahead.
Well, that went on a lot longer than I wanted it to, and covered some thoughts completely off topic. For those wanting the most concise sumarry possible: Dell making Macs can never happen. Intel switch a necessary evil. Yay Dev Tools and Cocoa frameworks!
