Coastal seas worldwide suffer from increasing human impact. One of the most severe environmental threats is excessive nutrient pollution from land, which causes oxygen depletion and harmful algal blooms. In 2018, the semi-enclosed Baltic Sea was determined to contain the largest hypoxic area among the world's coastal seas, with a size equal to the Republic of Ireland. In this study, ensemble modelling was used to investigate whether climate change will intensify hypoxia in the Baltic Sea and whether nutrient load abatement strategies would counteract this scenario. We analysed the largest ensemble of scenario simulations for the Baltic Sea currently available (including different boundary conditions) and estimated the magnitude of various sources of uncertainty. The results showed that natural variability was a larger source of uncertainty than previously considered. The earliest time and appropriate location to detect a trend above the background noise were estimated. A significant decrease in hypoxia can be achieved by further reductions in nutrient loads implemented in combination with existing measures. Natural variability makes future projections of hypoxia in the Baltic Sea uncertain, but nutrient load reductions can reduce its severity, shows a large range of future scenarios downscaled from an ensemble of Earth System models.