Or MIPS64. Or Alpha. Or Sparc64. Or even, heaven forbid, what Intel, HP, CPQ and others (even Dell) hilariously used to call "Industry Standard 64" (which most of the world called Itanic).

Volume is key, as other comments have already noted. But these days there's no volume without volume applications, and no volume applications without a volume OS. Once upon a time, in the early days of NT, there was the Advanced RISC Computing consortium (Alpha, MIPS, PPC), and for a while, there was a prospect of a critical mass actually building up. But Billco effectively killed that opportunity when he reverted NT to x86-only, and the concept of RISC as an x86 alternative in the volume market died.

The Linux community could have continued to support alternatives to x86, but why would they, especially after AMD64 showed that 64bit didn't have to be totally incompatible with IA32 (unlike Itanium)? So the "general purpose" Linux market is also largely x86 (routers, STBs, phones, etc are a distinct market).

For the foreseeable future, the Wintel ecosystem *is* the volume market, for better or worse. They have the monopoly (even more so since Apple went x86), and monopolists don't give up their monopoly without a struggle.

As far as parallelism in software goes: there are two times when most people's PCs have a chance to exploit parallelism: (1) when background-printing something with something else in the foreground (2) when doing something in the foreground while a full-system virus scan is running in the background. Anything else, such as full-system indexing in the background, is likely an irrelevance; most of the things most PCs do most of the time simply cannot be usefully parallelised, and even the complex things some PCs do some of the time cannot usefully be parallelised in general (and unless they are CPU-limited, which few are, there's no benefit to parallelising them anyway).

Parallelisation is only new and shiny in the desktop/volume market, it's old hat in the real computer market, and it's clear to those with clue that parallelisation is not going to work any miracles in the desktop market. But it fills column inches, so that's OK.