Background The immune reconstitution after initiation of highly active antiretroviral therapy (HAART) among HIV-infected individuals substantially affects patients' prognosis. However, the dynamic characteristics and predictors of reconstitution outcome remain unclear. Methods In this study, the HIV/AIDS patients with sustained virological suppression (viral load < 50 copies/ml) after HAART were enrolled. The patients were subgrouped into immunological non-responders (INRs) (< 200 cells/mu l), immunological inadequate responders (IIRs) (200 500 cells/mu l) and immunological responders (IRs) (> 500 cells/mu l) according to the CD4 cell count after two-year HAART. The immune reconstitution data based on the CD4(+) and CD8(+) cell counts with 8-year follow-up were collected for analysis. Results The CD4(+) cell counts in the immunological responders (IRs) were significantly higher than in the immunological non-responders (INRs) and immunological inadequate responders (IIRs) (P < 0.001). The overall CD4(+) cell count and CD4/CD8 ratio in the IRs increased faster than the IIRs and INRs. The CD4(+) cell count growth at 0.5 year and 1 year after HAART in the IRs was significantly higher than the IIRs and INRs. The ROC curve demonstrated that 1 year CD4(+) cell count had the highest predictive value, with the best cut-off value of 188 cells/mu l, the predictive sensitivity was 81.0%, the predictive specificity was 85.2%, false positive rate was 14.8%, false negative rate was 19.0%, positive predictive value (IR) was 63.0%, negative predictive value (INR) was 93.5%. Conclusions Taken together, our findings suggest that early initiation of HAART can reduce the immune reconstitution failure. The combination of baseline CD4(+) cell count and baseline CD4/CD8 ratio may serve as a valid predictor of immune reconstitution prognosis after HAART.