Late dinuclear transition-metal (especially group 10 and 11) homoleptic carbonyl complexes are elusive species and have so far not been isolated. A typical example is the 30-electron species [Ni2(CO)5], the structure and bonding of which is still debated. We show that, by using the AlCp* ligand (isolobal to CO), it is possible to isolate and fully characterize [Ni2(AlCp*)5] (1), which inspired us to revisit by DFT calculations, the bonding situation within [Ni2L5] (L = CO, AlCp*) and other isoelectronic species. The short Ni-Ni X-ray distance in 1 (2.270 Å) should not be attributed to the existence of a typical localized triple-bond between the metals, but rather to a strong through-bond interaction involving the three bridging ligands via their donating lone pairs and accepting π* orbitals. In contrast, in the isostructural 32-electron [Au2(AlCp*)5] (2) cluster an orbital with M-M antibonding and Al...Al bonding character is occupied, which is in accordance with the particularly long Au-Au distance (3.856 Å) and rather short Al...Al contacts between the bridging ligands (2.843 Å). This work shows that, unlike late transition-metal [M2(CO)x] species, stable [M2(AlCp*)x] complexes can be isolated, owing to the subtle differences between CO and AlCp*. We propose a similar approach for rationalizing the bonding in the emblematic 34 electron species [Fe2(CO)9].