They're really convinced the PC as we know it is dying.
What will fail to kill it is a failure to commodotize.
I don't think that last sentence actually has a useful meaning. Maybe what you meant is that the failure of attempts to commodotize will keep the PC alive?
Something like a C64 or a Raspberry Pi could be sold as a commodity-- sealed case, no servicable parts, all soldered to the board surface-mount. But when you've got grandma spending $249 for the entire PC, and gamers spending twice that on one video card, the market is still far from ready for that.
One problem for your theory is a new invention called "the video game console." Another is that a market has to have a certain critical size to survive - that some people are willing to spend $500 on a video card isn't enough to call them into existence; you have to have a pretty good number.
One problem with predicting the future for PC gaming hw is that figures show PC gaming growing dramatically.. but a lot of this is social gaming, played via flash on Facebook. Otoh is that figures for total games sold by digital distribution are basically made up - and a very large part of the total.
The "thin clients for everyone" vision some people love just shifts the problem-- instead of large, complex boxes, you have heavy-capacity networking and big servers, along with many failure points.
Ok: this is just wrong. Networks and servers are redundant - if a physical server you are relying on goes down, another one steps in. The network is a net, not a chain, so there are many paths, not one.
Also: net bandwidth is rising, and I know that VCs and game companies have been looking at game servers that do all the rendering for a client and send it over the net. Servers are local-ish, so latency is supposed to be acceptable. Obviously this does away with any possibility for piracy - customers never own games, they just buy playing time.
Bottom line: at the moment, the future of gaming is anybody's guess.