Abstract Background The possible mediating role of cardiovascular disease (CVD) in the relationship between alcohol use disorders (AUD) and the risk of early-onset (<age 65) and late-onset (≥age 65) dementia lacks formal investigation. Methods Using linked Finnish national register data, a population-based cohort study of 262,703 dementia-free Finnish men and women aged 40 + at baseline (December 31, 1999) was established. AUD and CVD in 1988–2014, and incident dementia in 2000–2014 were identified from Finnish Hospital Discharge Register and/or Drug Reimbursement Register. Causal association and mediation were assessed using mediational g-formula. Results AUD was associated with a substantial increase in the risk of early-onset dementia in both men (hazard ratio: 5.67, 95% confidence interval: 4.37–7.46) and women (6.13, 4.20–8.94) after adjustments for confounding; but the elevated risk for late-onset dementia was smaller (men: 2.01, 1.80–2.25; women: 2.03, 1.71–2.40). Mediational g-formula results showed that these associations were causal in men with no mediation by CVD as the virtually identical total effect of AUD (early-onset: 5.26, 3.48–7.48; late-onset: 2.01, 1.41–2.87) and direct effect of AUD (early-onset: 5.24, 3.38–7.64; late-onset: 2.19, 1.61–2.96) were found with no indirect effect via CVD. In women, the results were similar for late-onset dementia (total effect: 2.80, 1.70–4.31; direct effect: 2.92, 1.86–4.62) but underpowered for early-onset dementia. Conclusion AUD increased dementia risk, particularly the risk of early-onset dementia. This elevated risk of dementia associated with AUD was not mediated by CVD. Clinicians should consider the increased risk of dementia in the management of middle-aged and older adults with a history and/or current AUD.