Sepsis – syndrome of infection complicated by organ dysfunction – is responsible for over 750 000 hospitalisations and 200 000 deaths in the USA annually. Despite potential nutritional benefits, the association of diet and sepsis is unknown. Therefore, we sought to determine the association between adherence to a Mediterranean-style diet (Med-style diet) and long-term risk of sepsis in the REasons for Geographic Differences in Stroke (REGARDS) cohort. We analysed data from REGARDS, a population-based cohort of 30 239 community-dwelling adults age ≥45 years. We determined dietary patterns from a baseline FFQ. We defined Med-style diet as a high consumption of fruit, vegetables, legumes, fish, cereal and low consumption of meat, dairy products, fat and alcohol categorising participants into Med-style diet tertiles (low: 0–3, moderate: 4–5, high: 6–9). We defined sepsis events as hospital admission for serious infection and at least two systematic inflammatory response syndrome criteria. We used Cox proportional hazard models to determine the association between Med-style diet tertiles and first sepsis events, adjusting for socio-demographics, lifestyle factors, and co-morbidities. We included 21 256 participants with complete dietary data. Dietary patterns were: low Med-style diet 32·0 %, moderate Med-style diet 42·1 % and high Med-style diet 26·0 %. There were 1109 (5·2 %) first sepsis events. High Med-style diet was independently associated with sepsis risk; low Med-style diet referent, moderate Med-style diet adjusted hazard ratio (HR) 0·93 (95 % CI 0·81, 1·08), high Med-style diet adjusted HR=0·74 (95 % CI 0·61, 0·88). High Med-style diet adherence is associated with lower risk of sepsis. Dietary modification may potentially provide an option for reducing sepsis risk.