another way of saying the same thing is that the maths used in the physical models doesn't "embed" the universe in anything. there's no need for an external reference. you can do everything internally, consistently. it's quite neat, actually.
that doesn't mean that there "really is" nothing. just that nothing extra is needed for the theory, and since nothing extra can be observed, it's simpler to not have anything.
in other words - you can mathematically describe a universe that expands, but doesn't need anything to expand into. since that fits observations (within reason) and is simpler than a model that needs something extra "outside" that also needs explaining (and how would you explain that?!), it's preferred.