Hepatocellular carcinoma (HCC) is a highly malignant tumor that responds very poorly to existing therapies, most probably due to its extraordinary inter‐ and intra‐tumor molecular heterogeneity. The modest therapeutic response to molecular targeted agents underlines the need for new therapeutic approaches for HCC. In our study, we took advantage of well‐characterized human HCC cell lines, differing in transcriptomic subtypes, DNA mutation and amplification alterations, reflecting the heterogeneity of primary HCCs, to provide a preclinical evaluation of the specific heat shock protein 90 (HSP90) inhibitor AUY922 (luminespib). Indeed, HSP90 is highly expressed in different tumor types, but its role in hepatocarcinogenesis remains unclear. Here, we analyzed HSP90 expression in primary human HCC tissues and evaluated the antitumor effects of AUY922 in vitro as well as in vivo. HSP90 expression was significantly higher in HCC tissues than in cirrhotic peritumoral liver tissues. AUY922 treatment reduced the cell proliferation and viability of HCC cells in a dose‐dependent manner, but did not do so for normal human primary hepatocytes. AUY922 treatment led to the upregulation of HSP70 and the simultaneous depletion of HSP90 client proteins. In addition, in a cell type‐dependent manner, treatment induced either both caspase‐dependent β‐catenin cleavage and the upregulation of p53, or Mcl‐1 expression, or NUPR1 expression, which contributed to the increased efficacy of, or resistance to, treatment. Finally, in vivo AUY922 inhibited tumor growth in a xenograft model. In conclusion, HSP90 is a promising therapeutic target in HCC, and AUY922 could be a drug candidate for its treatment.