The triglyceride-glucose (TyG) index could serve as a convenient substitute of insulin resistance (IR), but epidemiological evidence on its relationship with the long-term risk of mortality is limited. Participants from the National Health and Nutrition Examination Survey during 1999-2014 were grouped according to TyG index (<8, 8-9, 9-10, >10). Cox regression was conducted to compute the hazard ratios (HRs) and 95% confidence interval (CI). Restricted cubic spline and piecewise linear regression were performed to detect the shape of the relationship between TyG index and mortality. A total of 19,420 participants (48.9% men) were included. On average, participants were followed-up for 98.2 months, and 2,238 (11.5%) and 445 (2.3%) cases of mortality due to all-cause or cardiovascular disease were observed. After adjusting for confounders, TyG index was independently associated with an elevated risk of all-cause (HR, 1.10; 95% CI, 1.00-1.20) and cardiovascular death (HR, 1.29; 95% CI, 1.05-1.57). Spline analyses showed that the relationship of TyG index with mortality was non-linear (All non-linear < 0.001), and the threshold value were 9.36 for all-cause and 9.52 for cardiovascular death, respectively. The HRs above the threshold point were 1.50 (95% CI, 1.29-1.75) and 2.35 (95% CI, 1.73-3.19) for all-cause and cardiovascular death. No significant difference was found below the threshold points (All > 0.05). Elevated TyG index reflected a more severe IR and was associated with mortality due to all-cause and cardiovascular disease in a non-linear manner.