The development of low‐oxygen zones threatening marine life (hypoxia) occurs annually in multiple coastal regions of the world. The largest estuary of the continental United States, the Chesapeake Bay, typically has ≈10 km3 of water with dioxygen concentrations <3 mg L−1 in July. As numerical methods for refining model resolutions in targeted areas are becoming common, there is interest in assessing the feasibility of simulating coastal hazards such as hypoxia in Earth System Models (ESMs). These coupled models are typically not constrained by observations and thus likely to feature systematic biases in their land, atmosphere, or ocean components. This study relies on four numerical experiments to evaluate the sensitivity of Chesapeake Bay hypoxia to changes or biases in its external physical forcings. Hypoxia exhibits only a minor decrease (−1.6%) after reducing the Bay's terrestrial freshwater discharge by 9.5% (but keeping terrestrial nutrient loadings the same). Changes in freshwater discharges have their largest impact on hypoxia during one extreme event (−37% during 2011 tropical storm Lee). Similarly, changing oceanic conditions on the shelf or their temporal frequency impact hypoxia by only 5%–6%, indicating that the latter is predominantly dictated by local conditions. Although these results are promising from the perspective of ESMs, additional components of ESMs will need to be evaluated before general conclusions can be reached. We notably speculate that the Bay's hypoxia would exhibit higher sensitivity to other forcings not examined here, notably air temperatures and nutrient loadings. Plain Language Summary Many coastal embayments of the world experience the development of “dead zones” during the summer, defined by very low concentrations of oxygen in water. The dead zones negatively affect the ecosystem including commercially‐important fisheries. Such a dead zone can be found every summer in the Chesapeake Bay, the largest and most productive estuary in the continental United States. Although these phenomena are typically the focus of local or regional managers, recent improvements in computer capabilities raise the possibility that dead zones could be successfully represented in the same type of computer simulators used for century‐long projections of the Earth's climate. Such simulators would provide a long term projection for the severity of dead zones throughout the world. The present study examines whether a realistic representation of the Chesapeake Bay dead zone can be obtained using information from these global computer simulations. The results show that the representation of the dead zone is not substantially different when using the global data sets selected for this study. This positive outcome opens the door for additional tests that should focus on other components of the global simulators, notably air temperatures and nutrients inputs from the land to the coastal embayments. Key Points Chesapeake Bay hypoxia shows only a weak sensitivity to continental shelf conditions and variability Changes in shelf hydrography propagate throughout the Bay but surface temperatures are set by the meteorological forcing Short term hydrological events are not always captured well by global land models which can impact hypoxia substantially