Cassava is becoming increasingly important as an industrial raw material in China. However, an insufficient supply of cassava raw materials and the expanding demand for cassava in downstream-processing industries restricts the development of the cassava industry in China. This paper studies how to increase the scale of cassava planting and promoting cassava production efficiency using output-oriented Data Envelopment Analysis (DEA) modeling. Overall Technical Efficiency (OTE), Pure Technical Efficiency (PTE), and Scale Efficiencies (SE) of the cassava-production system in the major cassava production areas of China are calculated using Variable Returns to Scale (VRS). Results reveal that, in addition to the Guangdong province, the OTE of Guangxi, Hainan, Fujian, Yunnan, and Jiangxi is inefficient, with an OTE of less than 1. The largest cassava-planting province, Guangxi, has the lowest SE with 0.551. The PTE of Guangxi is close to the minimum with 0.344. The OTE of Guangxi is also the lowest among the five provinces with 0.190. This study also presents ways to improve production efficiency. Results reveal that Guangxi has a large ratio of transverse adjustment on average. In the Guangxi province, 1.70% of the service cost and 1.72% of the labor cost need to be eliminated to keep the current output scale. Meanwhile, service and labor costs need to be reduced by 3164.85 and 3209.92, respectively, to achieve the best production efficiency. Further industrialization and large-scale cassava cultivation, increased yield, and strengthened cooperation with the Association of South East Asian Nations (ASEAN) and Africa are suggested as policy options to improve the cassava system in China.