To address the association between exercise capacity and the onset of dementia, Alzheimer disease, and cognitive impairment. For 6104 consecutive veteran patients (mean ± SD age: 59.2±11.4 years) referred for treadmill exercise testing, the combined end point of dementia, Alzheimer disease, and cognitive impairment was abstracted from the Veterans Affairs computerized patient record system. After mean ± SD follow-up of 10.3±5.5 years, 353 patients (5.8%) developed the composite end point at a mean ± SD age of 76.7±10.3 years. After correction for confounders in multivariate Cox proportional hazards regression, higher age at exercise testing (hazard ratio [HR]=1.08; 95% CI, 1.07-1.09; P<.001), current smoking (HR=1.44; 95% CI, 1.08-1.93; P=.01), and exercise capacity (HR=0.92; 95% CI, 0.89-0.96; P<.001) emerged as predictors of cognitive impairment. Each 1-metabolic equivalent increase in exercise capacity conferred a nearly 8% reduction in the incidence of cognitive impairment. Meeting the recommendations for daily activity was not associated with a delay in onset of cognitive impairment (HR=1.07; 95% CI, 0.86-1.32; P=.55). Exercise capacity is strongly associated with cognitive function; the inverse association between fitness and cognitive impairment provides an additional impetus for health care providers to promote physical activity.