Plenty of structured, unstructured and semi-structured soil-related data have been accumulated in China, but they are from multiple sources with varied structures and stored dispersedly and disorderly. In the age of big data and artificial intelligence, in order to satisfy with the requirements of soil data mining and application, it is urgent to construct soil big data architecture. In this study, based on concepts and cutting-edge technologies of big data, data lake and data warehouse, the approaches to construct soil big data architecture were proposed, including technologies on soil-related data acquisition and preprocessing, unstructured data processing, algorithms and models. Then, soil data storage scheme was built through data lake and warehouse, and soil data sharing and serving schemes were realized with the support of data catalog management. Meanwhile, soil big data have been used in soil data service, intelligent control of contaminated sites and mining soil ecological patterns. Finally, the potentials of soil big data were discussed in the usage of construction of the database of third national soil census and prospects for the application in data-driven soil researches.