Epidemiologic evidence on metal mixtures and non-alcoholic fatty liver disease (NAFLD) is limited. We aimed to assess the relationship between multiple metal co-exposure and NAFLD among male adults in Northern China. We conducted a cohort-based case-control study with 648 NAFLD and 648 non-NAFLD males. Seven metal concentrations (calcium, copper, iron, magnesium, manganese, selenium, and zinc) were determined in the blood. We used logistic regression and restricted cubic splines (RCS) to estimate the associations between the single metal and NAFLD. The impact of metal mixtures was quantified by the environmental risk score (ERS) in the adaptive elastic-net regression, and the association with NAFLD was estimated by logistic regression. Age-adjusted RCS showed linear relationships between blood calcium, selenium, and NAFLD. Blood copper, iron, magnesium, and manganese were non-linearly associated with NAFLD. Single metal analysis observed significant relationships between calcium, copper, manganese, and NAFLD, with the adjusted odds ratio (95% confidence interval) for quartile 1 vs. quartile 4 of 1.99 (1.30, 3.05), 2.36 (1.52, 3.64), and 1.77 (1.22, 2.55), respectively. However, metal mixtures analysis revealed one squared term (copper [β = −0.146]) and five metal-metal interactions (calcium × copper [β = 0.200], copper × magnesium [β = 0.188], copper × selenium [β = 0.188], iron × magnesium [β = 0.143], magnesium × selenium [β = −0.297]) except the three main effects. Higher ERS indicated a higher risk for NAFLD when exposed to metal mixtures, with an adjusted odds ratio = 6.50 (95% confidence interval: 4.36–9.69) for quartile 4 vs. quartile 1. Mediation analysis suggested that 11.66% of the effect of ERS on NAFLD was suppressed by fasting blood glucose. Our results show that exposure to metal mixtures is associated with a higher risk for NAFLD than the single metal. Interactions between metals suggest the importance of balancing the various metals for health benefits. Prospective cohorts and mechanism studies need to confirm the findings.