Abstract Background Low serum triglyceride (TG) has been suggested as a predictor of mortality after cardiovascular disease. However, the relationship between the level of TGs and the outcome after stroke remains to be elucidated. We hypothesized that the influence of TG levels on post-stroke mortality varies according to stroke mechanism: cardioembolic (CE) vs. non-CE causes. Methods We prospectively enrolled a consecutive series of patients with first-ever acute ischemic stroke for 5 years (n=1067), and followed them until the end of 2007 to obtain information on mortality and cause of death. We divided the level of TG into the quartiles, and classified the patients into CE (n=226) and non-CE stroke groups (n=841). The influence of TG level on mortality (all-cause death and vascular death) was examined by univariate and multivariate analyses using Cox regression. Result All-cause death and vascular death rates showed inverse relationships to the quartiles of TG levels in all patients (p<0.001, both) and also in non-CE stroke group (p<0.001, both), but not in CE stroke group (p=0.17 and p=0.37, respectively). In the Cox-regression analysis, compared with the highest quartile, the adjusted hazard ratio (HR) of the lowest quartile for all-cause death was 2.58 [95% confidence interval (CI), 1.38–4.82] and that for vascular death was 3.50 (95% CI, 1.39–8.82) in non-CE stroke group. These same associations, however, were not significant in CE stroke group. Conclusion Our results indicate that low serum TG is an independent predictor of mortality after ischemic stroke brought on by non-CE causes.