In theory, you can use solid-state NMR to characterize most solids, as long as the solid in question contains elements with odd number of protons and/or neutrons. However, if the abundance of such isotopes is low, the NMR analysis will take very long.

The main practical difference between powder NMR and XRD is the scale of the interactions being measured: in NMR each atom's chemical shift is influenced by the closest neighboring atoms, while in XRD you measure long range ordering of atoms or molecules (typically between a couple of nanometers and microns).

For example, one advantage of using NMR instead of XRD to characterize a powder sample is the possibility of characterizing amorphous samples.  However, since NMR and XRD measure different interactions, I think it would be best to use both methods simultaneously if you can.

Agree with the above comment, I just like to add one more thing. When it comes to NMR analysis of inorganic compound, one must consider the paramagnetic or ferromagnetic property of the compound.

These two techniques are complementary to each other..      XRD  is very powerful tool for the structure analysis of perfect and imperfect solids including the short range order. In conjuction with  XFS attachements  it will be indispensable tool for the chemical and phase analysis.  NMR  with high temperature attachment will be powerful tool to study diffusion in  those solids contain species or their isotopes having non vanishing nuclear dipole moments (odd number of proton and/or neutron). Similarly using nuclear quadruple- electric field gradient interactions one can get the dislocation densities from the intensity measurements of NMR lines in heavily deformed metals ans alloys. Also the dislocation domain  size and the wall thickness may be obtained deformed samples for low stacking fault energy (SFE) metals by NMR line broadening analysis.

In addition, TEM may play very unique role by supplying the stacking fault width in low SFE alloys to calculate the density of the extended partial dislocations knowing  the  stacking faults probability by the XRD measurements. In the final stage the annealing kinetics of stacking faults can be track down by XRD to obtain the segregation kinetics of  solute atoms through their diffusion activation energy determinations.  See: our two published papers listed  at this platform.