I can't find the anonymous interview with the physics engine company that used to fund them until around 2004 or 5, but they weren't even as nice as Carmack or Persson. According to them, if the claims they fed potential investors are correct, it would require unheard of memory speeds and capacity, even with some kind of lossless instantaneous compression.
On the other hand, if Carmack and Persson are right, and it's just an advanced voxel engine (which it really looks like, along with some kind of tessellation gone insane terrain duplication based on their wider shots), it could probably be on the market in the next console generation, assuming they can convince anyone to sign on. Promising the moon and delivering black and white pictures of the desert is usually a good way to guarantee nobody wants your pictures. Even if they are really nice pictures.