Average usage of the CPU in the iPhone 4S must be pretty low under most circumstances. I don't see why we can't do it on the device now.It is possible.
Currently processing power and local storage are the hurdles for speech to text. It appears that the Galaxy Nexus can handle it. I just wonder what the CPU usage is and how much space those libraries take up.
nVidia also just introduced a new flagship variant of Tegra 3 capable of using whopping DDR3-1600 over the old DDR3-667. I expect the Nexus 7 to be using the entry Tegra 3 (T30L). On the mobile front you are still looking at single channel DDR3 and USB 2.0 for the majority of peripheral interfacing.
There was clamoring for Thunderbolt on the iPad or iPhone but what does it matter when the internal flash storage is still connected over USB 2.0? The next major improvement is going to be RAM speed and internal bus speed increases.
Shoot we have been able to dictate to our computers since 1997 (maybe even longer) with no requirement for internet connection. I would like to believe our iPhone 4S is as powerful as a 1997 PC.