Background Valproic acid (VPA) has shown beneficial effects in vitro against SARS-CoV-2 infection, but no study has analyzed its efficacy in the clinical setting. Methods This multicenter, retrospective study included 165 adult patients receiving VPA at the time of admission to hospital, and 330 controls matched for sex, age and date of admission. A number of clinical, outcome and laboratory parameters were recorded to evaluate differences between the two groups. Four major clinical endpoints were considered: development of lung infiltrates, in-hospital respiratory worsening, ICU admissions and death. Results VPA-treated patients had higher lymphocyte (P<0.0001) and monocyte (P = 0.0002) counts, and lower levels of diverse inflammatory parameters, including a composite biochemical severity score (P = 0.016). VPA patients had shorter duration of symptoms (P<0.0001), were more commonly asymptomatic (P = 0.016), and developed less commonly lung infiltrates (65.8%/88.2%, P<0.0001), respiratory worsening (20.6%/30.6%, P = 0.019) and ICU admissions (6.1%/13.0%, P = 0.018). There was no difference in survival (84.8%/88.8%, P = 0.2), although death was more commonly related to non-COVID-19 causes in the VPA group (36.0%/10.8%, P = 0.017). The cumulative hazard for developing adverse clinical endpoints was higher in controls than in the VPA group for infiltrates (P<0.0001), respiratory worsening (P<0.0001), and ICU admissions (P = 0.001), but not for death (0.6). Multivariate analysis revealed that VPA treatment was independently protective for the development of the first three clinical endpoints (P = 0.0002, P = 0.03, and P = 0.025, respectively), but not for death (P = 0.2). Conclusions VPA-treated patients seem to develop less serious COVID-19 than control patients, according to diverse clinical endpoints and laboratory markers.