Immunogenic cell death (ICD) is the kind of cell death that triggers the immune system. It affects several tumors, whereas its significance for prognosis in acute myeloid leukemia (AML) remains uncertain. AML categorization by cytogenetic variables is inaccurate. In addition, risk stratification of AML based on cytogenetics is imprecise. The data of AML patients were extracted from 4 databases, a total of 1,537 patients. Univariate and LASSO Cox regression analyses were conducted to construct an ICD risk signature (ICDRS). The ICDRS showed strong prognostic value for AML through Kaplan-Meier, Cox, ROC analyses and nomogram. Combining the ICDRS with the European LeukemiaNet (ELN) classification to redefine the risk stratification can better predict the prognosis of AML. Moreover, the ICDRS was examined to identify gene functional enrichment, immunological characteristics, drug susceptibility, and somatic mutation, which revealed considerable variations among different risk categories. We further validated the expression of ICDRS in the AML bone marrow microenvironment by single-cell RNA (scRNA) analysis. Ultimately, the functional role of CASP1 was proven in AML by a series of in-vitro experiments. Our study highlights the significant impact of ICDRS on AML, which may improve ELN risk classification, predict immune landscapes, and guide personalized therapy.