Studies have identified non-linear inverse relationships between physical activity and the risk of stroke. A U-shaped response has been observed between haemorrhagic stroke and physical activity. The objective of this study was to investigate the associations between physical activities on stroke. We used data from the E3N cohort study, a French prospective study of women initiated in 1990. From the women in the study, we included those without cardiovascular disease or cancer at baseline, resulting in 94,169 women. We assessed total physical activity in 1993, grouped as quartiles. Cox models adjusted for potential confounders were used to assess the relationship with stroke, considering cases until 2008. Splines were used to assess the shape of the response. Similarly, we then considered high and low-intensity physical activity grouped as tertiles. Among the included women, with a mean age of 51.2 ± 6.7 years, 592 cases of stroke were identified over an average follow-up time of 16.2 years. Total physical activity was associated with a lower stroke risk (HR(Q1-Q4) = 0.38 [0.30, 0.49]). An inverse relationship was observed between physical activity and all stroke sub-types. A non-linear (L-shaped) relationship was observed for all-stroke, and ischemic stroke, and a U-shaped response for sub-arachnoid and intracerebral haemorrhage. High-intensity activities were associated with a U-shaped response for haemorrhagic stroke types. Low-intensity activities were associated with a linear response for all stroke types. Our results support other observations that physical activity may reduce stroke risk. U-shaped responses were observed between physical activity and haemorrhagic stroke.