In fact, neither "picture" is right, they are pictures so that you can visualize a model of reality. There is a microscopic "vacuum" field between the atoms, and also between the electron and the nucleus. This microscopic, or rather femto/pico/nano-scopic field is both temporally and spatially strongly inhomogeneous and fluctuating, and there are two ways that its macroscopic average manifests itself. It is partly and mostly a bulk effect and partly and mostly a superficial surface effect. In macroscopic physics and engineering neither is more fundamental than the other. In the old days physicists and still today most engineers have preferred to think of fictitious charges as the source of magnetization because it is easier to visualize it; modern views are more leaning to view sources as being from currents but both macroscopic source concepts are fictitious mathematical constructs. Both can be traced to the so-called Helmholtz decomposition theorem, see this post.

In fact, neither "picture" is right, they are pictures so that you can visualize a model of reality. There is a microscopic "vacuum" field between the atoms, and also between the electron and the nucleus. This microscopic, or rather femto/pico/nano-scopic field is both temporally and spatially strongly inhomogeneous and fluctuating, and there are two ways that its macroscopic average manifests itself. It is partly and mostly a bulk effect and partly and mostly a superficial surface effect. In macroscopic physics and engineering neither is more fundamental than the other. In the old days physicists and still today most engineers have preferred to think of fictitious charges as the source of magnetization because it is easier to visualize it; modern views are more leaning to view sources as being from currents but both macroscopic source concepts are fictitious mathematical constructs. Both can be traced to the so-called Helmholtz decomposition theorem, see posts, and, and.