Preclinical studies on the role of erythropoietin (EPO) in bone metabolism are contradictory. Regeneration models indicate an anabolic effect on bone healing, whereas models on physiologic bone remodeling indicate a catabolic effect on bone mass. No human studies on EPO and fracture risk are available. It is known that fibroblast growth factor 23 (FGF23) affects bone mineralization and that serum concentration of FGF23 is higher in men with decreased estimated glomerular filtration rate (eGFR). Recently, a direct association between EPO and FGF23 has been shown. We have explored the potential association between EPO and bone mineral density (BMD), fracture risk, and FGF23 in humans. Plasma levels of EPO were analyzed in 999 men (aged 69 to 81 years), participating in the Gothenburg part of the population‐based Osteoporotic Fractures in Men (MrOS) study, MrOS Sweden. The mean ± SD EPO was 11.5 ± 9.0 IU/L. Results were stratified by eGFR 60 mL/min. For men with eGFR ≥60 mL/min (n = 728), EPO was associated with age (r = 0.13, p  < 0.001), total hip BMD (r = 0.14, p  < 0.001), intact (i)FGF23 (r = 0.11, p = 0.004), and osteocalcin (r = −0.09, p = 0.022). The association between total hip BMD and EPO was independent of age, body mass index (BMI), iFGF23, and hemoglobin (beta = 0.019, p  < 0.001). During the 10‐year follow‐up, 164 men had an X‐ray–verified fracture, including 117 major osteoporotic fractures (MOF), 39 hip fractures, and 64 vertebral fractures. High EPO was associated with higher risk for incident fractures (hazard ratio [HR] = 1.43 per tertile EPO, 95% confidence interval [CI] 1.35–1.63), MOF (HR = 1.40 per tertile EPO, 95% CI 1.08–1.82), and vertebral fractures (HR = 1.42 per tertile EPO, 95% CI 1.00–2.01) in a fully adjusted Cox regression model. In men with eGFR<60 mL/min, no association was found between EPO and BMD or fracture risk. We here demonstrate that high levels of EPO are associated with increased fracture risk and increased BMD in elderly men with normal renal function.