Would MS commit millions and an army of coders to create something that does the same thing as something that already exists and can use?
Or would they simply make what already exists work on all their intended vendors?
Reinvent bread?
It is almost always easier and cheaper to take high level design ideas and recode and incorporate concepts than to reuse someone else's existing code that is highly specific and very incomplete.
DX12 has to be fully backyards compatible with DX1 and cover a huge range of features and be vendor agnostic.
Making a threaded API is no rocket science. Nothing in mantle in revolutionary, it is just that DX and OGl have never been properly developed for years. Mantle is a logical re-write of the existing API and is little more than a proof of concept demonstration that works on a small subset of vendor specific GPUs. Even then there are still bugs to work out.