ObjectiveThis study examines the causal effects of varying exercise intensities on type 2 diabetes mellitus (T2D) through Mendelian randomization (MR) analysis, using genetic variants as instrumental variables.MethodsA two-sample MR analysis was performed, employing Inverse Variance Weighted (IVW) as the primary method, supported by weighted median, MR-Egger regression, MR-PRESSO, and MR robustness-adjusted contour scores. Data were obtained from the International Exercise Genetics Database (IEGD) and the Global Diabetes Research Consortium (GRC), encompassing over 150,000 individuals for exercise intensity and around 200,000 T2D patients and controls. SNPs linked to exercise intensity were selected based on genome-wide significance (P < 5 × 10^-8) and linkage disequilibrium criteria (distance >10,000 kb, r^2 < 0.001).ResultsThe IVW analysis suggested that high-intensity exercise might reduce T2D risk, but the association was not statistically significant (OR = 0.667, 95% CI = 0.104–4.255, P = 0.667). The wide confidence interval indicates uncertainty in the effect estimate. Low-intensity exercise showed no significant effect on T2D risk (OR ∼ 1.0). Sensitivity analyses, including weighted median and MR-Egger regression, confirmed no significant association between high-intensity exercise and T2D risk. The MR-PRESSO analysis found no significant outliers, and the global test for pleiotropy was non-significant (P = 0.455). Cochran’s Q test for heterogeneity in the IVW analysis was non-significant (Q = 12.45, P = 0.234), indicating consistency among SNP-derived estimates.ConclusionHigh-intensity exercise potentially reduces T2D risk, but the association is not statistically significant. Further research is needed to understand the complex relationship between exercise intensity and T2D.