We conducted a two-sample Mendelian randomization (MR) design to evaluate the causal relation between breast cancer and stroke. Genetic variants associated with breast cancer and stroke were both obtained from genome-wide association study summary data. The single nucleotide polymorphisms were selected as instrumental variables. Effect estimates were primarily evaluated using standard inverse variance weighted. Finally, sensitivity analyses were performed for the detection of potential pleiotropy and heterogeneity in the cause-effect evaluation. There was a causal association of ER-positive breast cancer (odds ratio = 0.11, 95% confidence interval: 0.08–0.16, P < .001), and ER-negative breast cancer (odds ratio = 1.04, 95% confidence interval: 1.00–1.07, P = .045) with stroke. MR-egger regression revealed that the cause-effect of ER-positive breast cancer (P < .001) is drove by the directional horizontal pleiotropy, while there was no directional pleiotropy in the cause-effect of ER-negative breast cancer (P = .82). Cochran Q-derived P-value from inverse variance weighted (P = .27) shown that the cause-effect of ER-negative breast cancer on stroke do not need to consider the effect of heterogeneity. In addition, the leave-one-out analysis showed no influential instruments driving the associations, suggesting robust results for all outcomes. The present MR study reveals that ER negative breast cancer increase the risk of stroke.