Understanding spatially correlated floods and modeling joint hazard associated with threshold exceedances across multiple locations is crucial for accurate estimation of continental-scale portfolio risk. This work uses a non-parametric copula-based spatial simulator to analyze peak floods across the United States to derive the first-of-its-kind continental portfolio risk estimates at the 10- and 100-year return levels. We find significant interdependence in floods across the nation, revealing the recurring pattern of extreme events affecting the Northeast, Central, West, and Northwest United States in the same year. The stochastic simulator effectively manages high-dimensional data and offers reliable uncertainty estimates for both spatially dependent floods and the aggregated flood losses at the continental level. El Niño-Southern Oscillation and Atlantic Multidecadal Oscillation are identified as statistically significant tele-connectors of aggregate loss. This research aims to advance the understanding of compound continental flood hazard and the potential large-scale climate teleconnections that lead to such compound floods.