Patients with cancer are at high risk of severe coronavirus disease 2019 (COVID-19), with high morbidity and mortality. Furthermore, impaired humoral response renders severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccines less effective and treatment options are scarce. Randomized trials using convalescent plasma are missing for high-risk patients. Here, we performed a randomized, open-label, multicenter trial ( https://www.clinicaltrialsregister.eu/ctr-search/trial/2020-001632-10/DE ) in hospitalized patients with severe COVID-19 (n = 134) within four risk groups ((1) cancer (n = 56); (2) immunosuppression (n = 16); (3) laboratory-based risk factors (n = 36); and (4) advanced age (n = 26)) randomized to standard of care (control arm) or standard of care plus convalescent/vaccinated anti-SARS-CoV-2 plasma (plasma arm). No serious adverse events were observed related to the plasma treatment. Clinical improvement as the primary outcome was assessed using a seven-point ordinal scale. Secondary outcomes were time to discharge and overall survival. For the four groups combined, those receiving plasma did not improve clinically compared with those in the control arm (hazard ratio (HR) = 1.29; P = 0.205). However, patients with cancer experienced a shortened median time to improvement (HR = 2.50; P = 0.003) and superior survival with plasma treatment versus the control arm (HR = 0.28; P = 0.042). Neutralizing antibody activity increased in the plasma cohort but not in the control cohort of patients with cancer (P = 0.001). Taken together, convalescent/vaccinated plasma may improve COVID-19 outcomes in patients with cancer who are unable to intrinsically generate an adequate immune response.