Obesity has been associated with cognition in observational studies; however, whether its effect is confounding or a reverse causality remains inconclusive. This study aimed to investigate the causal relationships of overall obesity, measured by body mass index (BMI), and abdominal adiposity, measured by waist-hip ratio adjusted for BMI (WHRadjBMI), and cognition across European and Asian populations using Mendelian randomization (MR) analysis. We used publicly available genome-wide association study (GWAS) summary data of European ancestry, including BMI (n = 322,154) and WHRadjBMI (n = 210,088) from the GIANT consortium, and cognition performance (n = 257,828) from the UK Biobank and COGENT consortium. Data for individuals of Asian ancestry were retrieved from Taiwan Biobank to perform GWAS for BMI (n = 65,689), WHRadjBMI (n = 65,683), and Mini-Mental State Examination (MMSE, n = 21,273). MR analysis was carried out using the inverse-variance weighted method for the main results. Further, we examined the overall pleiotropy by MR-Egger intercept, and detected and adjusted for possible outliers using MR PRESSO. No causal effect of BMI on cognition performance (beta 95% CI = 0.00 -0.07, 0.07, p value = 0.91) was found for Europeans; however, a 1-SD increase in WHRadjBMI was associated with a 0.07 standardized score decrease in cognition performance (beta 95% CI = -0.07 -0.12, -0.02, p value = 0.006). Further, no causal effect of BMI on MMSE (beta 95% CI = 0.01 -0.08, 0.10, p = 0.91) was found for Asians; however, a 1-SD increase in WHRadjBMI was associated with a 0.17 standardized score decrease in MMSE (beta 95% CI = -0.17 -0.30, -0.03, p = 0.02). In both populations, overall pleiotropy was not detected, and outliers did not affect the robustness of the main findings. This trans-ethnic MR study reveals that abdominal adiposity, as measured by WHR adjusted for BMI, impairs cognition, whereas weak evidence suggests that BMI impairs cognition.