The heart's geometry and its metabolic activity vary over the cardiac cycle. The effect of these fluctuations on phosphorus (31P) magnetic resonance spectroscopy (MRS) data quality and metabolite ratios was investigated. 12 healthy volunteers were measured using a 7 T MR scanner and a cardiac 31P-1H loop coil. 31P chemical shift imaging data were acquired untriggered and at four different times during the cardiac cycle using acoustic triggering. Signals of adenosine-triphosphate (ATP), phosphocreatine (PCr), inorganic phosphate (Pi) and 2,3-diphosphoglycerate (2,3-DPG) and their fit quality as Cramér-Rao lower bounds (CRLB) were quantified including corrections for contamination by 31P signals from blood, flip angle, saturation and total acquisition time. The myocardial filling factor was estimated from cine short axis views. The corrected signals of PCr and [Formula: see text]-ATP were higher during end-systole and lower during diastasis than in untriggered acquisitions ([Formula: see text]). Signal intensities of untriggered scans were between those with triggering to end-systole and diastasis. Fit quality of PCr and [Formula: see text]-ATP peaks was best during end-systole when blood contamination of ATP and Pi signals was lowest. While metabolite ratios and pH remained stable over the cardiac cycle, signal amplitudes correlated strongly with myocardial voxel filling. Triggering of cardiac 31P MRS acquisitions improves signal amplitudes and fit quality if the trigger delay is set to end-systole. We conclude that triggering to end-systole is superior to triggering to diastasis.