CD47 is overexpressed on the surface of many types of cancer cells, including T-cell acute lymphoblastic leukemia (T-ALL) cells. In this study, we have developed a diphtheria toxin-based bivalent anti-human CD47 immunotoxin (bi-CD47-IT) for the targeted therapy of CD47+ cancers using a unique diphtheria toxin-resistant yeast Pichia pastoris expression system. Bi-CD47-IT demonstrated compelling in vivo efficacy in multiple T-ALL cell line-derived xenograft (CDX) and patient-derived xenograft (PDX) mouse models. Bi-CD47-IT significantly prolonged the median survival of the tumor-bearing mice and highly effectively depleted the T-ALL blast cells in the peripheral blood, spleen, liver, bone marrow, brain, and spinal cord in the T-ALL CDX and PDX mouse models. Bi-CD47-IT cured 60% of tumor-bearing mice in a T-ALL Molt-4 CDX mouse model. Because CD47 is also expressed on normal tissues, including red blood cells and lymphocytes, specificity is a concern. We thus analyzed the in vitro binding avidity and hemagglutination of bi-CD47-IT in human red blood cells, finding no binding or hemagglutination. We further performed a toxicity study of bi-CD47-IT in humanized mice, which showed that bi-CD47-IT transiently depleted the human lymphocytes for ~4 weeks after the 10-day treatment. No clinical adverse events were observed. As a result, bi-CD47-IT appears to possess the "optimal" binding avidity, with effective binding to human CD47+ T-ALL tumor cells, no binding to human red blood cells and weak binding to human lymphocytes. We believe that bi-CD47-IT is a promising and safe therapeutic drug candidate for the targeted therapy of CD47+ cancers.