The association between dietary Fe intake and diabetes risk remains inconsistent. We aimed to explore the association between dietary Fe intake and type 2 diabetes mellitus (T2DM) risk in middle-aged and older adults in urban China. This study used data from the Guangzhou Nutrition and Health Study, an on-going community-based prospective cohort study. Participants were recruited from 2008 to 2013 in Guangzhou community. A total of 2696 participants aged 40–75 years without T2DM at baseline were included in data analyses, with a median of 5·6 (interquartile range 4·1–5·9) years of follow-up. T2DM was identified by self-reported diagnosis, fasting glucose ≥ 7·0 mmol/l or glycosylated Hb ≥ 6·5 %. Cox proportional hazard models were used to estimate hazard ratios (HR) and 95 % CI. We ascertained 205 incident T2DM cases during 13 476 person-years. The adjusted HR for T2DM risk in the fourth quartile of haem Fe intake was 1·92 (95 % CI 1·07, 3·46; Ptrend = 0·010), compared with the first quartile intake. These significant associations were found in haem Fe intake from total meat (HR 2·74; 95 % CI 1·22, 6·15; Ptrend = 0·011) and haem Fe intake from red meat (HR 1·86; 95 % CI 1·01, 3·44; Ptrend = 0·034), but not haem Fe intake from processed meat, poultry or fish/shellfish. The association between dietary intake of total Fe or non-haem Fe with T2DM risk had no significance. Our findings suggested that higher dietary intake of haem Fe (especially from red meat), but not total Fe or non-haem Fe, was associated with greater T2DM risk in middle-aged and older adults.