BACKGROUND: Breastfeeding may be associated with lower future risk of maternal type 2 diabetes (T2D). However, existing evidence is inconsistent and derived largely from studies in high-income countries. We assess the association of breastfeeding and breastfeeding duration with incident T2D among Chinese women. RESEARCH DESIGN AND METHODS: The prospective China Kadoorie Biobank recruited 512 724 adults from 10 localities across China between 2004 and 2008. During 11.8 years' follow-up, 12 011 cases of incident T2D were recorded among 283 855 female participants without prior diabetes. Cox regression was used to estimate adjusted HRs for incident T2D associated with ever breastfeeding, mean breastfeeding duration per child and lifetime breastfeeding duration. RESULTS: Overall, 98.6% of female participants were parous, among whom 97.2% reported ever breastfeeding, with mean lifetime breastfeeding duration and breastfeeding duration per child of 34.8 and 14.9 months, respectively. Among parous female participants, there was no clear association between ever breastfeeding and risk of incident T2D (adjusted HR 1.06 (95% CI 0.94 to 1.20)). A modest log-linear positive association was observed between lifetime breastfeeding duration and incident T2D among parous female participants who ever breastfed (1.01 (1.01 to 1.02) per 6 months longer breastfeeding), but this was attenuated after adjustment for parity (1.00 (0.99 to 1.01)). Mean breastfeeding duration per child was not associated with incident T2D (1.01 (0.99 to 1.02) per 6 months longer breastfeeding). CONCLUSIONS: In this population with almost universal childbearing and breastfeeding, there was no apparent association of ever breastfeeding or of breastfeeding duration with incident T2D.