Abstract Background Physical inactivity is a risk factor for type 2 diabetes (T2D) and dementia. However, it is unknown if physical activity (PA) intensity is associated with brain health in people with T2D. Therefore, this study aimed to determine (i) associations between PA intensity and step count with both cognition and brain structure and (ii) if apolipoprotein E-ε4 or insulin therapy modifies any associations. Methods Participants were people with T2D (n = 220; aged 55–86 years). An accelerometer worn over the right hip was used to obtain step count and moderate-to-vigorous PA (MVPA) averaged over 7 days. Cognition in 7 domains was obtained using a battery of neuropsychological tests. Brain structure was measured by Magnetic Resonance Imaging. Linear regression models were used to examine associations between step count, MVPA and each cognitive and Magnetic Resonance Imaging measure. Apolipoprotein E-ε4 × PA and insulin therapy × PA product terms were added to the models to examine effect modification. Results The mean age of participants was 67.9 (SD = 6.3). Higher step count was associated with greater hippocampal volume (β = 0.028, 95% CI = 0.005, 0.051). Insulin therapy modified the association between MVPA and attention–processing speed, such that associations were significant in people receiving insulin therapy (p for interaction = .019). There were no other significant associations. Conclusions Higher step count and greater time spent in MVPA may be associated with better hippocampal volume and attention–processing speed, respectively, in people with T2D. People with greater diabetes severity (receiving insulin therapy) may get more cognitive benefit from MVPA.