This study aims to explore the causal relationship between obstructive sleep apnea (OSA) and type 2 diabetes (T2D) using bidirectional Mendelian randomization (MR). The genetic data related to OSA were obtained from the FinnGen Biobank (Ncase=16, 761, Ncontrol=201, 194) in the Genome-wide association study (GWAS). Three single nucleotide polymorphism (SNP) were screened out as instrumental variable (IV) of OSA. The genetic data related to T2D were derived from a large Meta-analysis of GWAS (Ncase=62, 892, Ncontrol=596, 424), 114 SNP were selected as IV of T2D. Multiple MR methods were used for analysis and inverse variance weighted (IVW) was performed as main method. The sensitivity of MR analytic results was analyzed using MR-Egger and other methods, and the IV was evaluated using -value statistics. MR analysis showed that OSA was significantly associated with increased risk of T2D (=2.016, 95%: 1.185-3.429, <0.05). There was no significant relationship between T2D and OSA risk (=1.030, 95%: 0.980-1.082, =0.238). There was heterogeneity in both-way results (OSA➝T2D, =1.808×10; T2D➝OSA, =1.729×10), and no horizontal pleiotropy (OSA➝T2D, =0.477; T2D➝OSA, =0.349). IV of OSA and T2D-selected in the study were strong instrumental variables ( statistics of OSA=20.543; statistics of T2D=30.117). Our results supported that OSA was a risk factor for T2D, but T2D had no significant impact on the incidence of OSA. Blood glucose monitoring and diabetes screening in OSA patients might be beneficial to the early detection and intervention of T2D.