Stroke is the predominant cause of death worldwide. We aimed to investigate the association of serum beta-2 microglobulin ( 2M) concentrations with risk of stroke and all-cause mortalities in a cohort study. Overall, 4914 U.S. adults (mean age = 63.0 years, 44.3% male) were recruited from the National Health and Nutrition Examination Survey (NHANES Ⅲ). During a median follow-up of 19.4 years, 254 stroke deaths and 3415 all-cause deaths were identified by the National Center for Health Statistics. The associations of 2M with stroke and all-cause mortalities were investigated by using weighted Cox proportional hazard regression models. 2M was positively associated with stroke and all-cause mortality in unadjusted models and multivariable-adjusted models. The multivariable HR (95% CI) for stroke mortality in Q5 VS Q1 of serum 2M concentrations was 3.45 (1.33-8.91; p for trend = 0.001) and that for all-cause mortality was 3.95 (3.05-5.12; p for trend 0.001). In subgroup analyses, the association of 2M and stroke mortality did not vary by different levels of sociodemographic and general stroke risk factors (p interaction 0.05). In addition, the magnitude of positive association between 2M with all-cause mortality did vary by age, ratio of family income to poverty, smoking status, and history of hypertensive (p interaction 0.05). Our findings suggest that support that 2M may be a marker of stroke and all-cause mortality, which provides a new perspective for the study of cerebrovascular health and long-term survival in the future.