Incorrect. All speech processing is done through Apple's servers. There is no special "chip" needed. I had Siri running perfectly fine on my iPhone 4 until I upgraded to a 5.
Processing what you meant is done on Apple's servers, but processing the speech itself is mostly done on your device.
The Audience chip in the iPhone 4S was used to ignore noise in the recording.
I'd bet that in the same noisy spot, an iPhone 4S/iPhone 5 would have much better voice recognition than an iPhone 4 hacked to run Siri.
In a quiet room, recognition rates would most likely be the same.
That's also the reason why it is very likely that Siri on Mac will run on any Mac capable of running OS X 10.9 - because they aren't meant to be used in noisy environments (like railway platforms) and seldom will be used in those environments - so there's no need for special technology to boost the voice recognition rates in newer hardware.
And of course - voice recognition (the Siri dictation part) is already implemented, so unless Siri on Mac can do really amazing stuff that requires more computing power than available on a 2012 MacBook Air, there's no need for new hardware.
And IF it could do such things that require more hardware performance - i would gladly update, because that would change the whole way to interact with a desktop/notebook computer.
But i doubt that we'll see something like that in Siri's first OS X incarnation.