Type 2 diabetes (T2D) is a prevalent disease that arises from complex molecular mechanisms. Here, we leverage T2D multi-ancestry genetic associations to identify causal molecular mechanisms in an ancestry- and tissue-aware manner. Using two-sample Mendelian Randomization corroborated by colocalization across four global ancestries, we analyze 20,307 gene and 1,630 protein expression levels using blood-derived