Simple sphere packings of metallic atoms are generally assumed to exhibit highly delocalized bonding, often visualized in terms of a lattice of metal cations immersed in an electron gas. In this Article, we present a compound that demonstrates how covalently shared electron pairs can, in fact, play a key role in the stability of such structures: Mo2Cu(x)Ga(6-x) (x ≈ 0.9). Mo2Cu(x)Ga(6-x) adopts a variant of the common TiAl3 structure type, which itself is a binary coloring of the fcc lattice. Electronic structure calculations trace the formation of this compound to a magic electron count of 14 electrons/T atom (T = transition metal) for the TiAl3 type, for which the Fermi energy coincides with an electronic pseudogap. This count is one electron/T atom lower than the electron concentration for a hypothetical MoGa3 phase, making this structure less competitive relative to more complex alternatives. The favorable 14 electron count can be reached, however, through the partial substitution of Ga with Cu. Using DFT-calibrated Hückel calculations and the reversed approximation Molecular Orbital (raMO) method, we show that the favorability of the 14 electron count has a simple structural origin in terms of the 18 - n rule of T-E intermetallics (E = main group element): the T atoms of the TiAl3 type are arranged into square nets whose edges are bridged by E atoms. The presence of shared electron pairs along these T-T contacts allows for 18 electron configurations to be achieved on the T atoms despite possessing only 18 - 4 = 14 electrons/T atom. This bonding scheme provides a rationale for the observed stability range of TiAl3 type TE3 phases of ca. 13-14 electrons/T atom, and demonstrates how the concept of the covalent bond can extend even to the most metallic of structure types.