Motivated by experimental and theoretical interest in realizing multipolar
orders in $d$-orbital materials, we discuss the quantum magnetism of $J\!=\!2$
ions which can be realized in spin-orbit coupled oxides with $5d^2$ transition
metal ions. Based on the crystal field environment, we argue for a splitting of
the $J\!=\!2$ multiplet, leading to a low lying non-Kramers doublet which hosts
quadrupolar and octupolar moments. We discuss a microscopic mechanism whereby
the combined perturbative effects of orbital repulsion and antiferromagnetic
Heisenberg spin interactions leads to ferro-octupolar coupling between
neighboring sites, and stabilizes ferro-octupolar order for a face-centered
cubic lattice. This same mechanism is also shown to disfavor quadrupolar
ordering. We show that studying crystal field levels via Raman scattering in a
magnetic field provides a probe of octupolar order. We study spin dynamics in
the ferro-octupolar state using a slave-boson approach, uncovering a gapped and
dispersive magnetic exciton. For sufficiently strong magnetic exchange, the
dispersive exciton can condense, leading to conventional type-I
antiferromagnetic (AFM) order which can preempt octupolar order. Our proposal
for ferrooctupolar order, with specific results in the context of a model
Hamiltonian, provides a comprehensive understanding of thermodynamics, $\mu$SR,
X-ray diffraction, and inelastic neutron scattering measurements on a range of
cubic $5d^2$ double perovskite materials including Ba$_2$ZnOsO$_6$,
Ba$_2$CaOsO$_6$, and Ba$_2$MgOsO$_6$. Our proposal for exciton condensation
leading to type-I AFM order may be relevant to materials such as
Sr$_2$MgOsO$_6$.