Hepatocellular carcinoma (HCC) remains a severe health issue worldwide, especially in Asia. To date, molecular classifications proposed for the overall survival (OS) or recurrence-free survival (RFS) prediction of Asian HCC patients after hepatectomy are quite few and limited in clinical practice. Here, we established a molecular subtyping system for Asian HCC to facilitate prognosis evaluation. Firstly, differentially expressed genes (DEGs) (FDR [false discovery rate] <0.05) between different types of liver cancer and non-tumor tissue were screened. Among the DEGs solely between HCC and non-tumor samples, 185 genes simultaneously significantly associated with the OS and RFS were identified as HCC-characteristic genes. The molecular subtypes were developed based on the expression profiles of the 185 genes in the training dataset (TCGA [The Cancer Genome Atlas] dataset) using non-negative matrix factorization (NMF) clustering method. Patients were then classified into Subtype1 and Subtype2 groups denoting unfavorable and favorable clinical outcome respectively. The robustness and effectiveness of the molecular subtype was confirmed in another independent dataset (GSE14520) by the same clustering approach and Kaplan-Meier analyses. Moreover, functional prediction analysis revealed that the identified molecular signature was involved in chemotaxis, apoptosis and cell development associated pathways. Besides, the molecular signature was closely related to the clinical characteristics including TNM stage, preoperative alpha-fetoprotein (AFP) level and TP53 mutation. Furthermore, integration of the molecular subtype and TNM stage was demonstrated to improve risk stratification. Taken together, our molecular subtyping system exhibited great utility and potential in prognosis prediction and therapeutic decision making of Asian HCC patients.