Purpose Rituximab (R) plus CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) chemotherapy is the standard of care in previously untreated diffuse large B-cell lymphoma (DLBCL). Obinutuzumab (G) is a glycoengineered, type II, anti-CD20 monoclonal antibody. GOYA was a randomized phase III study that compared G-CHOP with R-CHOP in patients with previously untreated advanced-stage DLBCL. Methods Patients (N = 1,418) were randomly assigned to receive eight 21-day cycles of G (n = 706) or R (n = 712), plus six or eight cycles of CHOP. Primary end point was investigator-assessed progression-free survival (PFS). Results After median observation of 29 months, the number of investigator-assessed PFS events was similar between G (201; 28.5%) and R (215; 30.2%), stratified hazard ratio was 0.92 (95% CI, 0.76 to 1.11; P = .39), and 3-year PFS rates were 70% and 67%, respectively. Secondary end points of independently reviewed PFS, other time-to-event end points, and tumor response rates were similar between arms. In exploratory subgroup analyses, patients with germinal-center B cell-like subtype had a better PFS than did patients with activated B cell-like subtype, irrespective of treatment. Frequencies of grade 3 to 5 adverse events (AEs; 73.7% v 64.7%, respectively) and serious AEs (42.6% v 37.6%, respectively) were higher with G-CHOP compared with R-CHOP. Fatal AE frequencies were 5.8% for G-CHOP and 4.3% for R-CHOP. The most common AEs were neutropenia (G-CHOP, 48.3%; R-CHOP, 40.7%), infusion-related reactions (G-CHOP, 36.1%; R-CHOP, 23.5%), nausea (G-CHOP, 29.4%; R-CHOP, 28.3%), and constipation (G-CHOP, 23.4%; R-CHOP, 24.5%). Conclusion G-CHOP did not improve PFS compared with R-CHOP in patients with previously untreated DLBCL. AEs reported with G were consistent with the known safety profile. Biomarker analyses may help define a future role for G in DLBCL.