Assessing the risk of cropland abandonment in mountainous area is essential for ensuring food security. However, the current evaluation indicator system is incomplete, and the reliability analysis of the results is lacking. This paper proposes a spatially comprehensive assessment method of cropland abandonment risk based on AHP and PCA using multi-source data of land use, abandoned cropland, administrative area, population density, GDP, cropland soil information, and climate, and applies the method to Yunnan Province in China. Firstly, 14 indicators were selected from four aspects, including social factors, farming conditions, cropland quality, and climate factors, to construct a three-layer indicator system of “target-criteria-indicator”. Fuzzy membership was used to calculate the indicator layer, and then the PCA method was applied to calculate the first principal component for the criterion layer. Finally, the AHP method was used to calculate the weighted superposition of the criterion layer. The proposed method showed good reliability in the study areas. In Yunnan Province, 33.03% of cropland is at high risk of abandonment, 37.52% is at medium risk, and 29.45% is at low risk. The high risk area is mainly distributed in the Lancang River Basin and Honghe River Basin. Social factors were found to play the greatest role in the risk of cropland abandonment in Yunnan Province. Factors such as low population density, large shape coefficient of cropland, and low annual average precipitation were identified as the main drivers of cropland abandonment in this region. The research results of this paper can provide decision-making reference for preventing and controlling cropland abandonment in mountainous areas.