Background: Effective two-drug regimens could decrease long-term drug exposure and toxicity with HIV-1 antiretroviral therapy (ART). We therefore aimed to evaluate the efficacy and safety of a two-drug regimen compared with a three-drug regimen for the treatment of HIV-1 infection in ART-naive adults. Methods: We conducted two identically designed, multicentre, double-blind, randomised, non-inferiority, phase 3 trials: GEMINI-1 and GEMINI-2. Both studies were done at 192 centres in 21 countries. We included participants (≥18 years) with HIV-1 infection and a screening HIV-1 RNA of 500 000 copies per mL or less, and who were naive to ART. We randomly assigned participants (1:1) to receive a once-daily two-drug regimen of dolutegravir (50 mg) plus lamivudine (300 mg) or a once-daily three-drug regimen of dolutegravir (50 mg) plus tenofovir disoproxil fumarate (300 mg) and emtricitabine (200 mg). Both drug regimens were administered orally. We masked participants and investigators to treatment assignment: dolutegravir was administered as single-entity tablets (similar to its commercial formulation, except with a different film colour), and lamivudine tablets and tenofovir disoproxil fumarate and emtricitabine tablets were over-encapsulated to visually match each other. Primary endpoint was the proportion of participants with HIV-1 RNA of less than 50 copies per mL at week 48 in the intention-to-treat-exposed population, using the Snapshot algorithm and a non-inferiority margin of −10%. Safety analyses were done on the safety population. GEMINI-1 and GEMINI-2 are registered with ClinicalTrials.gov, numbers NCT02831673 and NCT02831764, respectively. Findings: Between July 18, 2016, and March 31, 2017, 1441 participants across both studies were randomly assigned to receive either the two-drug regimen (n=719) or three-drug regimen (n=722). At week 48 in the GEMINI-1 intention-to-treat-exposed population, 320 (90%) of 356 participants receiving the two-drug regimen and 332 (93%) of 358 receiving the three-drug regimen achieved plasma HIV-1 RNA of less than 50 copies per mL (adjusted treatment difference −2·6%, 95% CI −6·7 to 1·5); in GEMINI-2, 335 (93%) of 360 in the two-drug regimen and 337 (94%) of 359 in the three-drug regimen achieved HIV-1 RNA of less than 50 copies per mL (adjusted treatment difference −0·7%, 95% CI −4·3 to 2·9), showing non-inferiority at a −10% margin in both studies (pooled analysis: 655 [91%] of 716 in the two-drug regimen vs 669 [93%] of 717 in the three-drug regimen; adjusted treatment difference −1·7%, 95% CI −4·4 to 1·1). Numerically, more drug-related adverse events occurred with the three-drug regimen than with the two-drug regimen (169 [24%] of 717 vs 126 [18%] of 716); few participants discontinued because of adverse events (16 [2%] in the three-drug regimen and 15 [2%] in the two-drug regimen). Two deaths were reported in the two-drug regimen group of GEMINI-2, but neither was considered to be related to the study medication. Interpretation: The non-inferior efficacy and similar tolerability profile of dolutegravir plus lamivudine to a guideline-recommended three-drug regimen at 48 weeks in ART-naive adults supports its use as initial therapy for patients with HIV-1 infection. Funding: ViiV Healthcare.