Current coastal flood risk assessments fail to capture flood spatial dependence at large scales. In this paper, we develop the first global synthetic dataset of spatially-dependent extreme sea level events, by applying an existing conditional multivariate statistical model to 40-year global reanalysis sea levels. The resulting dataset contains 10,000 years of extreme events with realistic spatial dependence under current climate conditions. The benchmarking against reanalysis data demonstrates a high agreement, with a coefficient of determination (R2) of 0.96 for the mean event footprint sizes and a mean bias of −0.04 m for 1 in 50-year water levels. By comparing well-known historic events, we show that our approach can produce events with similar spatial characteristics. Our dataset enables the future development of spatially-dependent flood hazard maps for deriving accurate large-scale risk profiles, which can help yield new insights into the spatial patterns of coastal flooding and support coastal communities in devising effective management plans.