amk-aka-phantom, you've brought up a lot of great points, so I'll touch base upon them.
1. Building your own PC and cost
The Windows tax is what kills you. If you're building from scratch, you're looking at $100 just for the OEM license. Celeron 450 2.2ghz is $40. 2GB for $15. 500gb hdd for $40. LGA775 mobo for $50. DVD burner $20. Keyboard/mouse $15. $50 for the case and 250W supply. That's $330 before cost of Newegg's per-item shipping and then you have the labor of putting it together and then you don't have the luxury of a one-year warranty. If I was building a PC for someone, I could do it for a few bucks less than HP, or I could just send them to HP and not have to deal with being their tech support guy for life...
I've been building PCs since the 80286 era and even then, I was going enthusiast grade by running an NEC V20 instead of Intel's chip. I was overclocking when it meant desoldering and resoldering a new crystal on the motherboard. In those days, it didn't matter if you were building a high-end system or a low-end system, it was always considerably cheaper to build than it was to buy. The gains we can make by eating up the cost of labor, warranty/support, and not having to make a profit are offset by the ability of these companies to buy components in such bulk that their per-unit cost is cheaper than what we can get from Newegg which is the closest thing we have to getting "bulk rate" discounts.
You still do okay building your own PC because of component selection, but that isn't how it always was. In fact, if tier 1 OEMs had unlocked BIOS's for overclocking potential, the cost difference between buying and building would shrink even more.
2. I appreciate that you understand the importance of HP subsidizing enthusiast manufacturers. ASUS started off building circuit boards for Dell before they started building whole motherboards for Dell before they started building "just about everything." ASUS and Asetek are probably the two most obvious ones imho.
3. Samsung will never license WebOS. By licensing Android they've run into the scenario they have now. "Historically" the trick with Google was to rotate suppliers for their reference platform. HTC = Nexus One, Samsung = Nexus S, Motorola = Xoom. It keeps all of the tier 1 licensees happy and lets everyone get their turn at being the reference platform. Now that Google owns Motorola, it becomes a trickier situation.
Licensing WebOS puts them in the same situation as Android, only Android is 100x more popular. Buying WebOS outright gives them a lot of flexibility and value-added capabilities in terms of Bada. Some of Bada's strongest points are the advertising engine. But Samsung's experience with user interface isn't as strong as Palm's was and from a pure "let's ignore the app situation" perspective, WebOS is far more usable than Android. It'll all boil down to price. If WebOS was $100K, they'd buy it right away. If WebOS is $100M, they're not going to jump on it just yet.
Guys like Qualcomm got their graphics expertise by outright buying ATI's mobile graphics division, which itself was an outright buy of BitBoys. That means that San Diego based Qualcomm has a Finnish division focused exclusively on the graphics technology. For Samsung to buy out WebOS, there needs to be a Samsung OS division based in California. It makes sense from a technical standpoint if you know the strengths/weaknesses of WebOS and the strengths/weaknesses of Bada and believe that there is opportunity for integration of Bada's strengths into WebOS (more likely than the other way around). The financial standpoint is something that only can be determined if you have inside info on the numbers being thrown around.
3. Re: Office on ARM.
The mistake that everyone had was thinking that Google was just a search engine. Or how the iPad would just be a big iPhone. Remember that ARM was engineered as a desktop CPU for Acorn "back in the day." I'm saying stuff like Office on a ARM-powered notebook and desktop. If it had honest-to-goodness 100% Microsoft Office support, I think you'd see a lot more casual users skipping the PC buying experience.
And my argument is that economies of scale help the enthusiasts too.
4. On power supplies.
I've been writing about power supplies before it was vogue to talk about power supplies.
http://www.firingsquad.com/guides/power_supply/default.asp
There are two areas where we've run into trouble when it comes to PSUs. The first is overclocking. As enthusiasts we're so used to overclocking that you're bummed if you're not running a Nehalem or Sandy Bridge beyond 4GHz. We take that for granted. Same with the GPU.
With the Firebird, HP was pulling off SLI'd 9800M GS's which probably put your load at about 120W. The Core 2 Quad 9550 had a TDP of 95W like every other Intel CPU, but at 12MB L2 and 45nm, it was actually closer to that than not.
http://www.xbitlabs.com/articles/cpu/display/power-consumption-overclocking_16.html#sect0
With a GTX560i at 170W, we're eating up another 50W or so. But the Sandy Bridges are almost half the power consumption of the Bloomfield core first-gen i7's at the same clock speeds.
http://www.xbitlabs.com/articles/cpu/display/core-i7-2600k-990x_12.html
Then by going with SSD instead of a platter drive, going with modern motherboard chipsets instead of the old nForce stuff, you're making incremental gains. Then, consider the fact that our power supplies are much more efficient today than they were before. That means for the same rated PSU, you're actually getting better performance. The Seasonic X is my preferred design right now.
It's crazy, I know, but it's definitely within striking distance provided that you had a fully custom system and PSU where the rails and the design was balanced just right. We just go with over-engineered PSUs because it lets us overclock without second thought about whether or not a PSU will be the issue, because it'll let us choose any GPU we want, and because it just lets you forget about it.
It would require a lot of careful design and execution, but it's 100% possible with everything being custom manufactured.
5. The PSU form factor is not beneficial for modern system cooling. That's why we had stuff like motherboards that are rotated 90 degrees or the old Lian Li PC-V line which tried to use the PSU fan to cool HDDs (but would only work for a front-flow design as opposed to a top/down).
I'm going to say that you just don't know what's possible, system-cooling wise, if we weren't tied down the legacy form factor.
6. HP tends to source their consumer PSUs from Hipro which is actually a reasonable manufacturer. The problem is that on the mainstream CPUs, they provide the bare minimum power required for a stable system with the possible configured options.
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=148
In the HP systems I've seen at work, etc. They haven't died within a year. The question is if those systems have a big buildup of dust and if ventilation is a problem. (Getting back to my "PSUs suck because of legacy design choices soap box").
7. HP enthusiast products? Not a lot. Articooler. Primary OEM of Asetek water cooling. Major OEM of Asus motherboards and GPUs. Developer of Voodoo Blackbird (which started before the Voodoo buyout). Voodoo Firebird. That's probably it. From a workstation enthusiast product, I don't think there's a better workstation than a Z800 + Dreamcolor display.
But it's not about buying HP products directly. It's what a major OEM does for enthusiasts indirectly when that OEM is a company like HP.
The firebird and blackbird were exactly what you were talking about having a good chassis, and good BIOS, and a good motherboard. They got too cocky and tried to push the limits. Just as crazy as you think a 350W i7-2600K and GTX560 Ti might be, a modern day Firebird would probably go for something like that with a external Seasonic X style PSU. It ended up being too expensive for consumers.
1. Cloud computing.
Do you use Gmail, Hotmail or Yahoo? Remember it's the average consumer that we're talking about.
2. Tablets/smartphones
You keep assuming that ARM = tablet/smartphones. I see the big companies looking to cut costs by giving "average users" poor performing hardware that seems to do what they need. You can complain about Apple dumbing down the public, but sadly, that's the public that exists today.
3. PCs will keep getting faster and faster. There's no question about that. But our software isn't keeping up and average users aren't seeing tangible differences between fast/slow CPUs. There was a time when a faster CPU actually meant that your web pages loaded faster. That was thanks to bad browsers, crazy table based layouts, and the state of hardware at the time. Average home user is still happy with a Core 2 Duo. When tablets reach Core 2 Duo performance, will they care that the Ivy Bridge CPUs are super fast? Not unless there are applications that convince them to do so. Outside of games and digital imaging/video, I'm not sure what CPU-intensive or GPU-intensive tasks exist for normal users.
"Just like they said, if a person already has a 2-year-old PC and it can do everything s/he needs, there will be no upgrade, and it makes sense."
Exactly. And the danger/concern is that with today's software/hardware, the upgrade cycle will be even slower. Maybe it'll be like a car. With Windows 8 being more efficient than Windows 7, do you think consumers will be able to do "everything s/he needs" with an i5-2500K in 2015?
Then, while this is true, do you think the *R&D costs* for next-generation GPUs and CPUs are going down or going up? What about R&D costs for better motherboards?
If the PC mainstream market slows down, that's less R&D play money for the companies making enthusiast products. That's what I'm worried about, and that's why the HP future concerns me.