Background: Homocysteine has been identified as a potential risk factor for stroke, cerebral small-vessel diseases (SVD), and dementia. Objective: The present study aimed to investigate the predictive value of homocysteine levels on incident dementia while simultaneously controlling for MRI findings and vascular risk factors. Methods: Within a Japanese cohort of participants with vascular risk factors in an observational study, we evaluated the association between baseline total homocysteine (tHcy) levels (per 1 μmol/L and the tertile of tHcy), the prevalence of MRI-findings at baseline, and incident all-cause dementia. Baseline brain MRI was used to determine SVD (lacunas, white matter hyperintensities, and cerebral microbleeds [CMBs]) and atrophy (medial-temporal lobe atrophy and bicaudate ratio). Logistic regression analyses were used to estimate the cross-sectional association between tHcy and each of MRI findings. Cox proportional hazards analyses were performed to estimate the longitudinal association between tHcy and dementia. Results: In the 643 subjects (age: 67.2 ± 8.4 years, male: 59% ; education: 12.9 ± 2.6 years), multivariable analyses adjusted for several potential confounders, including estimated glomerular filtration rate (eGFR) and intima-media thickness, showed that highest tHcy tertile was associated with lacunas, CMBs, and strictly deep CMBs. During the mean 7.3-year follow-up (range: 2–13), 47 patients were diagnosed with dementia (Alzheimer’s disease: 24; vascular dementia: 18; mixed-type: 3; other: 2). After adjusting for age, gender, APOE ɛ4, education, BMI, MMSE, hypertension, cerebrovascular events, eGFR, and MRI-findings, tHcy level (hazard ratios [HR]: 1.08, p = 0.043) and the highest tertile of tHcy (HR: 2.50, p = 0.047) for all-cause dementia remained significant. Conclusions: Our results provide additional evidence of tHcy that contributes to increased susceptibility to dementia risk.