Running more hardware requires more programming which equals higher costs which breaks down to a higher price, cry me a river.
I'm not crying, if anything I'm laughing at those spending that much money because Microsoft makes them think 8 CPUs need a radically different processor scheduler than 2 CPUs, whether it's really there or not is anyone's guess as anyone that sees the code surely signs an NDA to never talk about it.
Obviously I'll use Linux for an example, with the 'O(1)' process scheduler by Ingo Molnar it takes the exact same amount of time to schedule 10 takes on 1 CPU as it does 10,000 tasks on 32 CPUs, the 4 CPU sweet spot for Linux is gone. And as an added bonus the new scheduler is simpler and easier to maintain from what I read.
Same thing goes for memory, the main thing is getting past 4G because of the 32-bit limits of the most common processors, once those are gone the rest is relatively simple. The trick is dealing with poorly designed devices or ISA devices that can't DMA to/from high memory, but that's a driver issue and the maintainers of those drivers know about those problems already.
And now Linux has things like NUMA support, hot-swappable CPUs and even a new kernel module system and through all of it the maintenance cost isn't growing exponentially for most people maintaining kernel code (except for the IDE people, but that seems to be because each manufacturer seems to treat the ATA spec a little differently, once the new IDE taskfile core is done it'll be much simpler to add special cases for 'bad' chipsets and drives) because most of it is inconsequential. Apply a new process scheduer and all the drivers still work just fine, same for the VM subsystem (we even saw this in action in 2.4.x for better or worse) and many other things.
I'm sure the only real code differences between Server and Advanced server consist of things like:
#ifdef advsrv
#define max_mem 8196
#define max_cpu 4
...
Of course Microsoft is free to market and sell their products anyway they choose, but that doesn't mean you're not paying for a placebo.