Climate models exhibit large biases in sea ice area (SIA) in their historical simulations. This study explores the impacts of these biases on multimodel uncertainty in Coupled Model Intercomparison Project phase 5 (CMIP5) ensemble projections of 21st century change in Antarctic surface temperature, net precipitation, and SIA. The analysis is based on time slice climatologies in the Representative Concentration Pathway 8.5 future scenario (2070–2099) and historical (1970–1999) simulations across 37 different CMIP5 models. Projected changes in net precipitation, temperature, and SIA are found to be strongly associated with simulated historical mean SIA (e.g., cross-model correlations of r = 0.77, 0.71, and −0.85, respectively). Furthermore, historical SIA bias is found to have a large impact on the simulated ratio between net precipitation response and temperature response. This ratio is smaller in models with smaller-than-observed SIA. These strong emergent relationships on SIA bias could, if found to be physically robust, be exploited to give more precise climate projections for Antarctica.