Mantle is not open. There is a reason why we have high level apis on windows and other plattforms. It's a vendor specific api like Cuda.
It will be specific for AMD's cards, but not for the same reason that something like PhysX is limited to having all NV Cards (unless you used the driver trick that resulted from a mistake on NV's part). NV locked down PhysX by choice as a business decision (and an acceptable one).
Mantle, being close to the metal, will be specific to the hardware it is targeting. In this regard, it's closer to why assemblers aren't interchangeable for different architectures - they target specific functions provided by the hardware (arch/ISA), so they are in effect locked in by the hardware inherently. There is no going around the fact that if you want to code as close to the metal as possible, there is no way you can make it cross-ISA/cross-hardware, unless the hardware from different archs follow a set of standards in order to make such a thing possible (but then again, if that were the case, they probably wouldn't be completely different archs). Unlike PhysX that's been proven to work even if you don't have an NV card as your main card, because the limitation was artificial (but, in my opinion, not an unexpected business move from NV to make their own products more desirable).
Not that I disapprove of NV's locking down of PhysX. I think it's their call to make, and it's a fine call either way (lock it down or not). I don't have much of an opinion over it, and honestly if it were my product, and I want to sell more GPU's, I might have ended up with the same decision.
I'm only sending this clarification because it's important to be aware of the subtleties in why something like Mantle will always be vendor specific - because any initiative like Mantle (i.e., providing close to the metal access to hardware)
will always end up specific to that hardware, so locked to whoever provides that hardware.
And providing the option for Mantle is a good thing. No one is stopping devs from using traditional DX or OpenGL, and they all certainly will continue to use it (no PC dev will abandon DX/OpenGL to go solely on Mantle, that's crazy talk). But seeing as to how AMD has the same hardware in consoles, and that hardware already has close to metal access for devs, it makes sense to also provide that level of access to PC game devs, since the desktop cards from AMD will share the same arch. Win-win scenario.
EDIT: Wow this thread updates fast! I'd like to quote this one that appeared as I was composing:
Mantle is NOT vender independent.
Mantle is NOT AMD specific.
Mantle is GCN specific.
As Communism so succintly put it: Mantle, being close to metal, is inherently hardware (arch) specific - therefore GCN. Not locked by vendor through tricks. Not an artificial barrier. Not locked through a business decision. It is simply the nature of the beast whenever "close to the metal" access is your objective, whether you are talking of DSP's, GPU, CPU, etc.