As storm surge caused by tropical cyclones (TCs) can be immensely damaging, it is of the utmost importance to understand how this risk may change in a warming climate. While numerical models are capable of capturing surge with high accuracy, it remains computationally difficult to simulate a large enough sample of storm surge events to characterize the tails of the risk distribution. To address this gap, we present new results gathered from the efficient DeepSurge neural network storm surge model, as applied to nearly one million synthetic TC events in the North Atlantic basin downscaled from historical and future climate simulations under a strong warming scenario. By coupling with simulations from the Torrent flood inundation model, we evaluate how projected changes in surge height return periods relate to actual inundated area. This framework enables the quantification of climate-driven storm surge risk—particularly low-probability, high-impact events—in terms of its impact to human life, wellbeing, and infrastructure.