Background: Smoking is an established risk factor for type 2 diabetes mellitus (T2DM), yet whether smoking cessation confers equivalent protection across all age groups remains unclear.Methods: We examined the association between smoking status and incident T2DM in the KoGES prospective cohort (n=7,594; 14-year follow-up), stratified by age at baseline (≤50 vs >50 years), and replicated findings in the UK Biobank (n=127,866). To investigate biological underpinnings, DMP analyses were performed using genome-wide DNA methylation data from an independent epigenetic cohort, stratified by smoking status and age.Findings: In KoGES, former smokers aged ≤50 years showed complete normalisation of T2DM risk (HR 0·91, 95% CI 0·69–1·22), indistinguishable from never smokers. In contrast, former smokers aged >50 years retained persistently elevated risk equivalent to current smokers (HR 1·41, 95% CI 1·09–1·83; current smokers HR 1·44, 95% CI 1·12–1·84). This age-dependent divergence was replicated in the UK Biobank (younger former smokers HR 1·12, p=0·082; older former smokers HR 1·19, 95% CI 1·11–1·27). Epigenetic analyses identified 16 smoke-like loci—including AHRR, NEGR1, IL18RAP, and GNAS—that retained smoking-associated methylation patterns in older former smokers but normalised in younger former smokers.Interpretation: The protective effect of smoking cessation against T2DM is fundamentally age-dependent. Cessation before age 50 restores diabetes risk to never-smoker levels, whereas cessation after age 50 leaves a persistent epigenetic legacy that may sustain metabolic risk. These findings support targeted diabetes screening for older former smokers regardless of cessation duration.