Convergent dynamic functional connectivity studies have demonstrated their potential as a hallmark for capturing the impairments in brain function associated with Alzheimer's disease (AD) and mild cognitive impairment (MCI). However, understanding whole-brain dynamics patterns remains limited, hampering the understanding of cognitive impairment and symptomatology for AD and MCI. An energy-landscape analysis was conducted to investigate brain dynamics across seven large-scale networks in 516 normal controls (NC), 404 AD patients, and 441 MCI participants from a multi-center cohort. This method identified major brain states and quantified their size, duration, and transitions. In AD and MCI, transitions between these major states were excessively frequent, state durations were abnormal, and brain state sizes were enlarged. Furthermore, direct transitions between major states were significantly negatively correlated with cognitive ability and structural characteristics. This study has revealed aberrant brain dynamics in large-scale networks among patients compared to NC, suggesting that patients experience less stable states and more frequent transitions. The brain dynamic-cognition and dynamic-structure associations indicate that the dynamics of brain states could serve as a critical biological endophenotype of AD. These findings provide new insights into understanding and addressing brain network dynamics within AD and MCI.Copyright © 2025. Published by Elsevier Inc.