Thechemical bonds between a transition metal (M) and oxygen (O) are of major importance in catalysis, mineralogy, biology, and astrophysics, and an adequate theoretical description of these bonds is thus highly needed. This paper establishes that despite recent debate on its accuracy for transition-metal systems, CCSD(T) is an excellent benchmark standard for transition-metal oxide interactions, with errors approaching those of experiment. We conclude this from a study of all 60 M–Oand M+–O bond dissociation enthalpies (BDEs) of the 3d, 4d, and 5d metals, constituting a balanced data set in terms of dq configurations that also enable an assessment of the trend chemistry in oxygen’s ability to bind to transitionmetals. The BDEs decrease toward the right of the transition-metal series, with humps at groups 4, 5 and 8, 9. The linear trend followsthe increasing electronegativity when going from the left to the right, whereas the humps are caused by differential occupation of bonding δ-orbitals and antibonding π-orbitals. We show that the BDEs correlate strongly with oxophilicity and energies of metal surface chemisorption (R2 = 0.81–0.89); i.e., the local M–O bond dominates the energetics of transitionmetals reacting with oxygen. Therefore, theoretical studies of oxygen-involvingtransition-metal chemistry should emphasize the accuracy of the local M–O bonds. A “bottom-up” approach to theoretical catalysis may thus produce more accurate trend predictions of relevance to, for example, catalyst design. Finally, our analysis explains the large differences in chemisorption of oxygen on metal surfaces asprimarily caused by the metal electronegativity relative to oxygen, defining the strength of the polar covalent bonding, and secondarily caused by d-orbital net bonding.