Objective Data regarding low maternal haemoglobin concentration and severe maternal morbidity (SMM) are limited and potentially biased. This study evaluated the relation between early maternal haemoglobin concentration and SMM or maternal mortality. Design Population-based cohort study. Setting Ontario, Canada, in a public healthcare system. Population 737 393 births with a routine outpatient haemoglobin measured at a calculated gestational age of 2-16 weeks. Methods The relation between early-pregnancy outpatient blood haemoglobin concentration and each study outcome was expressed as adjusted relative risks (aRR) and absolute risk differences (aRD), with 95% confidence intervals (CI), generated by modified Poisson regression. Main outcome measures The primary outcome was SMM or maternal mortality, from 23 weeks' gestation to 42 days postpartum. Results The mean (SD) haemoglobin concentration was 126.9 (9.3) g/l. Overall, SMM or death occurred in 13 514 pregnancies (1.8%). Relative to a haemoglobin level of 125-129 g/l, the aRR was 1.07 (95% CI 1.02-1.13) and aRD (0.09%, 95% CI 0.01-0.18) at 120-124 g/l; aRR 1.31 (95% CI 1.17-1.46) and aRD 0.47% (95% CI 0.24-0.69) at 105-109 g/l; and aRR 4.53 (95% CI 3.59-5.72) and aRD 5.94% (95% CI 4.12-7.76) at <90 g/l. In all, 5961 women (0.8%) required red cell transfusion, with significantly higher risks at all haemoglobin concentrations below 125-129 g/l, peaking at a haemoglobin level <90 g/l (aRR 11.82, 95% CI 9.30-15.03). Conclusion There is a gradual increase in the risk of SMM or death, as well as red cell transfusion, starting from the lower level of the normal range of haemoglobin of non-pregnant women.