Aims Interventional trials provide evidence for a beneficial effect of reduced dietary sodium intake on blood pressure. The association of sodium intake with heart failure which is a long‐term complication of hypertension has not been examined. Methods and results Hazard ratios HRs, 95% confidence interval (CI) of heart failure comparing quintiles of estimated 24 h urinary sodium excretion (USE) were calculated in apparently healthy men (9017) and women (10 840) aged 39–79 participating in the EPIC study in Norfolk. During a mean follow‐up of 12.9 years, 1210 incident cases of heart failure occurred. Compared with the reference category (128 mmol/day ≤USE ≤148 mmol/day), the top quintile (USE ≥191 mmol/day) was associated with a significantly increased hazard of heart failure (1.32, 1.07–1.62) in multivariable analysis adjusting for age, sex, body mass index, diabetes, cholesterol, social class, educational level, smoking, physical activity, and alcohol consumption, with a marked attenuation (1.21, 0.98–1.49) when further adjusting for blood pressure. The bottom quintile (USE ≤127 mmol/day) was also associated with an increased hazard of heart failure (1.29, 1.04–1.60) in multivariable analysis without relevant attenuation by blood pressure adjustment (1.26, 1.02–1.56), but with substantial attenuation when adjusting for interim ischaemic heart disease and baseline C‐reactive protein levels and exclusion of events during the first 2 years (1.18, 0.96–1.47). Conclusion We demonstrate a U‐shaped association between USE and heart failure risk in an apparently healthy middle‐aged population. The risk associated with the high range of USE was attenuated after adjustment for blood pressure, whereas the risk associated with the low range of USE was attenuated after adjustment for pre‐existing disease processes.