The inter-patient variability of tumor proteomes has been investigated on a large scale but many tumors display also intra-tumoral heterogeneity regarding morphological and genetic features. It remains largely unknown to what extent the local proteome of tumors intrinsically differs. Here, we used hepatocellular carcinoma as a model system to quantify both inter- and intra-tumor heterogeneity across human patient specimens with spatial resolution.  We defined proteomic features that distinguish neoplastic from the directly adjacent non-neoplastic tissue, such as decreased abundance of NADH dehydrogenase complex I. We then demonstrated the existence of intra-tumoral variations in protein abundance that re-occur across different patient samples, and affect clinically relevant proteins, even in the absence of obvious morphological differences or genetic alterations. Our work demonstrates the suitability and the benefits of using mass spectrometry based proteomics to analyze diagnostic tumor specimens with spatial resolution. Data are available via ProteomeXchange with identifier PXD007052.