Abstract Purpose Hemoglobin (Hb) transfusion thresholds are established in intensive care units. A restrictive transfusion threshold (Hb 70–75 g/l) is recommended in septic patients, and a liberal transfusion threshold (Hb 90 g/l) for cardiogenic shock. It is unclear whether these historically adopted transfusion thresholds meet the challenges of individual patients. Methods We evaluated microvascular flow index (MFI) and proportion of perfused vessels (PPV) in the sublingual microcirculation with CytoCam-IDF microscopy and near-infrared spectroscopy (NIRS). A study team-independent, treating intensivist assigned a total of 64 patients to 1 of 2 two transfusion thresholds, 43 patients to the Hb 75 g/l threshold and 21 patients to the Hb 90 g/l threshold, at a surgical intensive care unit. We performed microcirculatory measurements 1 h before and 1 h after transfusion of 1 unit of red blood cells. Results Microcirculatory flow variables correlated negatively with pre-transfusion flow variables (ΔMFI: ρ = − 0.821, p 2.84, PPV > 88%) showed a deteriorated microcirculation after red blood cell transfusion. An impaired microcirculation improved after transfusion. At both transfusion thresholds, approximately one third of the patients showed an initially impaired microcirculation. In contrast, one third in every group had good microcirculation above the cutoff variables and did not profit from the transfusion. Conclusion The data suggest that the established transfusion thresholds and other hemodynamic variables do not reflect microcirculatory perfusion of patients. Blood transfusion at both thresholds 75 g/l and 90 g/l hemoglobin can either improve or harm the microcirculatory blood flow, questioning the concept of arbitrary transfusion thresholds.