I think the main reason why there is no huge demand for audio hardware is that it's perfectly possible to do render 20-30 three-dimensional sources in realtime in software, in CD quality (and, without totally killing the CPU). The difference between 20 sources, 200 sources, and 2000 sources is very small, if audible at all. Therefore it is conceivable to get away with fewer.
Monitor speakers and headsets are often of embarrassingly low quality too, so even if the sound isn't the best possible quality, a lot of people won't notice at all (and they'll not notice the difference between the most expensive soundcard and the onchip one, either).
It is, on the other hand, not trivially possible to do a similar thing with 3D graphics (not at present-day resolutions, and not with state-of-the-art quality, anyway). The difference between 20, 200, and 2000 objects on screen is immediately obvious. Displays are usually quite good, so the difference between good graphics and bad graphics is immediately obvious, too.
That doesn't mean that OpenAL is not being developed at all, however. The OpenAL-Soft implemention, which is kind of a de-facto standard (as compared to the dinosaur reference implementation) undergoes regular updates and implements several useful self-made extensions.