The physiological changes associated with type 2 diabetes mellitus begin before disease onset, yet few have examined the incidence of cancer both before and after diabetes onset. We examined the temporal relationship between diabetes and breast cancer risk. Breast cancer risk was assessed in a retrospective cohort study using patient data from the Marshfield Clinic electronic medical record including 5423 women who developed diabetes between 1 January 1995 and 31 December 2009 (reference date) and 26 346 nondiabetic women matched by age, smoking history, residence, and reference date. Breast cancer risk was assessed before and after reference date, adjusting for matching variables, BMI, insurance status, and comorbidities. Primary outcomes included hazard ratio (HR) and number of women needed to be exposed to diabetes for one additional person to be harmed - that is, develop breast cancer (NNEH). HR for breast cancer before diabetes diagnosis was 1.16 (95% CI 1.03-1.31, P=0.0150) and NNEH was 99 at time of diabetes onset. HR for breast cancer after diabetes diagnosis was not significant at 1.07 (95% CI 0.90-1.28, P=0.422), and NNEH was 350 at 10 years post diabetes onset. Diabetic women are at the greatest increased risk of breast cancer near the time of diabetes diagnosis. The comparative NNEH increased shortly after diagnosis and as the duration of diabetes increased. Breast cancer risk appears to be increased during the prediabetes phase, waning after diagnosis, raising important issues regarding timing of breast cancer prevention interventions in women with diabetes.