Sarcoendoplasmic reticulum Ca-ATPase (SERCA) is a transmembrane pump that plays an important role in transporting calcium into the sarcoplasmic reticulum (SR). While calcium (Ca) binds SERCA with micromolar affinity, magnesium (Mg) and potassium (K) also compete with Ca binding. However, the molecular bases for these competing ions' influence on the SERCA function and the selectivity of the pump for Ca are not well-established. We therefore used in silico methods to resolve molecular determinants of cation binding in the canonical site I and II Ca binding sites via (1) triplicate molecular dynamics (MD) simulations of Mg, Ca, and K-bound SERCA, (2) mean spherical approximation (MSA) theory to score the affinity and selectivity of cation binding to the MD-resolved structures, and (3) state models of SERCA turnover informed from MSA-derived affinity data. Our key findings are that (a) coordination at sites I and II is optimized for Ca and to a lesser extent for Mg and K, as determined by MD-derived cation-amino acid oxygen and bound water configurations, (b) the impaired coordination and high desolvation cost for Mg precludes favorable Mg binding relative to Ca, while K has limited capacity to bind site I, and (c) Mg most likely acts as inhibitor and K as intermediate in SERCA's reaction cycle, based on a best-fit state model of SERCA turnover. These findings provide a quantitative basis for SERCA function that leverages molecular-scale thermodynamic data and rationalizes enzyme activity across broad ranges of K, Ca, and Mg concentrations.