In the CELESTIAL trial for patients with advanced hepatocellular carcinoma (HCC), cabozantinib showed improved survival compared with placebo but comes at a price. We aimed to investigate the cost-effectiveness of cabozantinib for sorafenib-resistant HCC from the payer's perspective of the USA, UK and China. We developed Markov models to simulate the patients pre-treated with first-line sorafenib following the CELESTIAL trial. Quality-adjusted life-years (QALYs) and incremental cost-effectiveness ratio (ICER) were calculated for the treatment with cabozantinib or best supportive care. The list price for drugs was acquired from the Red Book, the British National Formulary, West China hospital and reported literature. Adverse events, utilities weights, and transition likelihood between states were sourced from the published randomized phase III trial. A willing-to-pay threshold was set $150 000/QALY in the USA, $70 671/QALY (£50 000/QALY) in the UK and $26 481/QALY (3x GDP per capita) in China. Deterministic and probabilistic sensitivity analyses were developed to test the models' uncertainty. In the base case, treatment with cabozantinib increased effectiveness by 0.13 QALYs, resulting in an ICER vs best supportive care of $833 497/QALY in the USA, $304 177/QALY in the UK and $156 437/QALY in China. The models were most sensitive to assumptions about transitions to progression with both cabozantinib and best supportive care, the utility associated with being progression free. These results were robust across a range of scenarios and sensitivity analyses, including deterministic and probabilistic analyses. Cabozantinib at its current cost would not be a cost-effective treatment option for patients with sorafenib-resistant HCC from the payer's perspective in the USA, UK or China. Substantial discounts are necessary to meet conventional cost-effectiveness thresholds. © 2019 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd.