<b><i>Introduction:</i></b> Currently, the effect of hemoperfusion on outcome in severe COVID-19 patients is still unknown. Therefore, we aimed to investigate the effects of early HA-330 hemoperfusion in severe COVID-19 patients. <b><i>Methods:</i></b> We conducted a single center, prospective cohort study on patients who were diagnosed with severe COVID-19 patients and admitted to ICU. Patients in hemoperfusion group (defined as patients who were treated with hemoperfusion therapy at least 3 sessions in combination with standard therapy) were compared with the control group (defined as patients who received standard treatment alone or received less than 3 sessions of hemoperfusion therapy). The primary outcome was daily sequential organ failure assessment (SOFA) scores. Secondary outcomes were all-cause mortality at 28 days, mechanical ventilator-free day, daily C-reactive protein (CRP), oxygenation (defined by PaO<sub>2</sub>/FiO<sub>2</sub> ratio), and severity score of lung infiltration on the chest X-ray (CXR RALE score). All outcomes were adjusted by regression analysis to reduce the confounders due to some difference in baseline characteristics. <b><i>Results:</i></b> A total number of 29 severe and critical COVID-19 confirmed patients were enrolled. Fifteen patients were defined as hemoperfusion group and 14 were control group. The median of CRP and SOFA score at the baseline (the day after severe pneumonia diagnosis or before hemoperfusion) in hemoperfusion and control groups were comparable, 96.79 mg/L and 87.3 mg/L, <i>p</i> = 0.53, 3.53 ± 0.99 versus 4.3 ± 1.89, <i>p</i> = 0.15, respectively. Clinical improvement associated with decreased SOFA score and improvement of CXR RALE score were found in hemoperfusion group compared to control group (<i>p</i> = 0.008 and <i>p</i> = 0.005, respectively). The 28-day mortality rate was significantly lower in hemoperfusion group compared to control group (6.67% vs. 85.71%, <i>p</i> < 0.001) and the adjusted hazard ratio of death was 0.017 (95% confidence interval = 0.008–0.351, <i>p</i> = 0.008). <b><i>Conclusions:</i></b> The addition of early HA-330 hemoperfusion to standard therapy improved severity of organ failure and might reduce the mortality rate. However, the results were affected by the baseline confounders and limited sample size.