With the growth of artificial intelligence (AI) technology, big data, “Internet+” and e-commerce have become an irreversible trend of the times. In the cross-border e-commerce industry chain, logistics demand forecasting is a key link. It requires precise prediction of the demand for each product in various warehouses in the future, to deploy the products to warehouses in various markets around the world in advance. This strategy can not only significantly shorten logistics time, but also greatly improve user experience, providing guarantees for the efficient operation of cross-border e-commerce. However, short-term logistics demand data often has complex characteristics such as non-stationarity, strong randomness, local variability, and nonlinearity, making precise prediction extremely challenging. This article proposes an intelligent system for forecasting cross-border logistics demand based on deep learning (DL) to address this challenge. This system quantitatively analyzes the uncertainty of logistics demand through DL technology, thereby achieving high-precision prediction of future logistics demand. The results indicate that our system performs well in prediction precision, providing strong technical support for optimizing cross-border e-commerce business.