This study aimed to detect the causal association of overweight and obesity on severe COVID-19 infection with respiratory failure through a two-sample Mendelian randomization (MR) method based on the genome-wide association studies datasets. All genome-wide association studies summary data of exposures and outcome used in this study were obtained from the IEU database derived from Europeans. The study mainly used the inverse variance weighted method to test causal relationship. Simultaneously, MR-PRESSO and MR-EGGER were used to detect the pleiotropy, and sensitivity analysis was performed using leave-one-out analysis. In the inverse variance weighted analyses, we found no causal association between obesity (e.g., OR = 1.15, 95% CIs = 0.96-1.37, P = .13 for obesity-ebi-a-GCST90000255), obesity subtypes (e.g., OR = 1.93, 95% CIs = 0.90-4.14, P = .10 for obesity and other hyperalimentation) as well as overweight (OR = 0.90, 95% CIs = 0.64-1.27, P = .54) and severe COVID-19 infection with respiratory failure. The findings showed no causal association between obesity or overweight and severe COVID-19 infection with respiratory failure. Further validation is needed regarding whether obesity or overweight is a risk factor for it.