OBJECTIVES: To compare the cost-effectiveness of atazanavir and boosted ritonavir (ATV/r) against lopinavir and boosted ritonavir (LPV/r), both with tenofovir and emitricitabine (TDF/FTC) backbone in Peru.  METHODS: We developed a Markov model from the perspective of the Peruvian Ministry of Health, for first and second-line treatment of HIV. We considered a 10 year time horizon, one-year Markov cycles, a 3% discount rate and quality adjusted life years as a measure of effectiveness. Costs were calculated in 2015 Peruvian Soles (S/.) and converted to U.S. Dollars (US$). We calculated an incremental cost-effectiveness ratio (ICER). To assess model uncertainty, we conducted one-way sensitivity analyses to evaluate individual cost drivers and probabilistic sensitivity analysis using Monte Carlo simulations. We considered a willingness to pay (WTP) threshold equal to the annual Peruvian Gross Domestic Product (GDP) per-capita (US$ 6 660). Fianlly, we calculated the net monetary benefit (NMB) considering a WTP range between zero and three times GDP per-capita. We used TreeAge 2015. RESULTS: The deterministic analysis showed that ATV/r is more effective and had higher costs than LPV/r, with an ICER of S/. 38 951 (US$ 11 423) for first-line treatment and S/. 163 660 (US$ 47 994) for second-line treatment. Probabilistic analysis showed that ATV/r would require a WTP greater than annual GDP per-capita in 67.1% and 53.2% of simulations in first and second-line treatment, respectively.  The one-way sensitivity analysis found that our model was robust to uncertainty. In first-line treatment, the NMB was higher for LPV/r until two-times GDP per-capita; after this threshold, ATV/r was more cost-effective. In second-line treatment the NMB was higher for LPV/r across WTP ranges.   CONCLUSIONS: Treatment with TDF/FTC/LPV/r is more cost-effective compared to TDF/FTC/ATV/r, both in naïve and treated patients. Nevertheless, ATV/r could be more cost-effective in first-line treatment if the WTP threshold was set at two-times annual GDP per-capita.