We conducted a cross-sectional study of NMR-derived HDL subclasses and alcohol intake among 2171 community-dwelling older adults with a large proportion of daily or near-daily alcohol consumers (44 %). We aimed to assess whether, in addition to increasing total HDL, alcohol may induce a beneficial shift in HDL particle size distribution. Participants were categorised based on reported alcohol intake (g per week) and on frequency (none, < 3 times/week, 3–4 times/week, ≥ 5 times/week). The association between alcohol intake and lipoprotein fractions was examined using sex-specific linear regression models adjusted for age, BMI, diabetes, current smoking, exercise and hormone therapy in women. There was a stepwise gradient with the highest weekly alcohol consumption associated with the highest total HDL size and greatest number of medium and large HDL particles, as well as higher total HDL concentrations (all P < 0·001); total small HDL did not differ. Alcohol–HDL size associations were similar in both sexes and did not differ by use of hormone replacement therapy in women. In conclusion, regular alcohol consumers had a higher number and percentage of large HDL particles than non-drinkers. These results suggest that one way that alcohol may decrease CVD is through potentially favourable changes in lipoprotein subclass composition.