Background: Mediterranean diet (MD) adherence has been associated with reduced risks of esophageal and gastric cancer (subtypes) in a limited number of studies. We prospectively investigated associations between MD adherence and risks of esophageal squamous cell carcinoma (ESCC), esophageal adenocarcinoma (EAC), gastric cardia adenocarcinoma (GCA), and gastric non-cardia adenocarcinoma (GNCA) in a Dutch cohort. Methods: Analyses were conducted using data from the 120852 participants of the Netherlands Cohort Study (NLCS), who were aged between 55 and 69 years at enrollment. Various MD scores, with and without alcohol, were calculated to estimate MD adherence. Using 20.3 years of follow-up, 133 ESCC, 200 EAC, 191 GCA, and 586 GNCA cases could be included in multivariable Cox regression analyses. Results: Of the investigated scores, the alternate Mediterranean diet score without alcohol (aMEDr) performed best. aMEDr was inversely associated with risks of GCA and GNCA in men and women. However, statistical significance was only reached in men [ptrend: 0.019 (GCA), 0.016 (GNCA)]. Furthermore, higher aMEDr values were significantly associated with a reduced ESCC risk in men [HRper two−point increment (95% CI) = 0.57 (0.41–0.80), ptrend = 0.013], but not in women (pheterogeneity = 0.008). There was no evidence of an association between aMEDr and EAC risk. Educational level was a significant effect modifier for the association between aMEDr and GNCA risk (pheterogeneity = 0.0073). Conclusions: Higher MD adherence was associated with reduced risks of ESCC, GCA, and GNCA in the NLCS. However, the decreased ESCC risk might be limited to men.