Given the growing interest in the metabolic heterogeneity of hepatocellular carcinoma (HCC) and portal vein tumour thrombus (PVTT). This study comprehensively analysed the metabolic heterogeneity of HCC, PVTT, and normal liver samples using multi-omics combinations. A single-cell RNA sequencing dataset encompassing six major cell types was obtained for integrated analysis. The optimal subtypes were identified using cluster stratification and validated using spatial transcriptomics and fluorescent multiplex immunohistochemistry. Then, a combined index based meta-cluster was calculated to verify its prognostic significance using multi-omics data from public cohorts. Our study first depicted the metabolic heterogeneity landscape of non-malignant cells in HCC and PVTT at multiomics levels. The optimal subtypes interpret the metabolic characteristics of PVTT formation and development. The combined index provided effective predictions of prognosis and immunotherapy responses. Patients with a higher combined index had a relatively poor prognosis (p <0.001). We also found metabolism of polyamines was a key metabolic pathway involved in conversion of metabolic heterogeneity in HCC and PVTT, and identified ODC1 was significantly higher expressed in PVTT compared to normal tissue (p =0.03). Our findings revealed both consistency and heterogeneity in the metabolism of non-malignant cells in HCC and PVTT. The risk stratification based on cancer-associated fibroblasts and myeloid cells conduce to predict prognosis and guide treatment. This offers new directions for understanding disease development and immunotherapy responses.