The changes will be industry-driven
I hear y'all who say sub-PC devices won't replace your PC any time soon; as a user, I'll still favor my dawn-of-7-era PC over such "toys". But as a PC builder and tech, I'm interested in the long view, and expect things to change.
Right now, sub-PCs remind me of the "home computer" era, before the ugliest hardware platform took over, benefitting as it did by an open hardware market where anyone could build interchangeable parts.
Like ye olde Atari, Amiga, Spectrum etc. they are tribal; you have to commit to Apple or BlackBerry or Android or Windows Mobile or Nokia, and having done so, are locked into that tribe's software and peripherals. And back in the home computer days, folks who made "real" computers (mainframes, UNIX minis) discounted the PC threat as being to weak to do more than act as a slightly less dumb terminal for the big iron of the day.
Right now, sub-PC devices are too costly to attract me away from desktop and laptop PCs. But if production volumes of these become the dog that wags the tail of component production, you can expect these to fall in cost, while PCs stay where they are - that, plus a Windows that favors the new devices, will see mobile PCs evolve towards the sub-PC in UI and hardware ergonomics. Once mass-produced tablets etc. offer greater economies of scale compared to PC components, this price gap will widen - possibly fueled by the "open hardware" effect of a Windows that will run on them (or maybe Android will do).
What I don't like, is having a communications vendor acting as a toll booth between me and my stuff (apps, data) "in the cloud". I also don't like the lack of hard boundaries that allow anyone with the magic password to wipe all your devices from the Internet, as happed to that Wired writer.
I think these figures are real and significant - I don't buy the notion that masses of folks are carrying over DDR3 RAM from old to new PCs. For one thing, PCs young enough to use DDR3 are prolly too new to be worth replacing yet; for a more obvious thing, very few people actually dig around at the component level or build their own PCs.
Other metrics to watch, are the shift in malware attention from PC to cross-platform surfaces (e.g. Java and Flash rather than Windows and IE), and the cost per Meg of comms bandwidth vs. local storage. The latter is illusionary, as bandwidth has to be consumed every time a file in "the cloud" has to be used, but a lot of folks will be too dim to notice that.
