The potential predictability of winter temperature in China from autumn Arctic sea ice anomalies is studied by examining and statistically modeling the large–scale interannual covariability between them on the basis of singular value decomposition analysis. It is demonstrated that an intimate relationship exists between September and October sea ice anomalies in the Eurasian Arctic and following winter temperature anomalies in China, except in the Tibetan Plateau. When the autumn sea ice anomalies decline in the Eurasian Arctic, above-normal pressure anomalies appear to prevail over the region from the Eurasian Arctic to Eastern Europe and Mongolia, and below-normal anomalies prevail over the mid-latitudes of Asia and Northwestern Pacific in the following winter. Consequently, the winter Siberian High and East Asian trough are both strengthened, favoring the southward invasion of high–latitude cold air masses and thus cold temperature anomalies in China. It is found that the Siberian High plays a crucial role in delivering effects of the autumn Arctic sea ice anomalies on winter temperature variability in China. Based on this evidence, a statistical model is established to examine the potential predictability of winter temperature anomalies in China by taking the autumn Arctic sea ice signals as a predictor. Validation shows considerable skill in predicting winter temperature anomalies over a large part of China, indicating a significant potential for improving winter climate prediction in China.