The gene that encodes de novo DNA methyltransferase 3A (DNMT3A) is frequently mutated in acute myeloid leukemia genomes. Point mutations at position R882 have been shown to cause a dominant negative loss of DNMT3A methylation activity, but 15% of DNMT3A mutations are predicted to produce truncated proteins that could either have dominant negative activities or cause loss of function and haploinsufficiency. Here, we demonstrate that 3 of these mutants produce truncated, inactive proteins that do not dimerize with WT DNMT3A, strongly supporting the haploinsufficiency hypothesis. We therefore evaluated hematopoiesis in mice heterozygous for a constitutive null Dnmt3a mutation. With no other manipulations, Dnmt3a+/- mice developed myeloid skewing over time, and their hematopoietic stem/progenitor cells exhibited a long-term competitive transplantation advantage. Dnmt3a+/- mice also spontaneously developed transplantable myeloid malignancies after a long latent period, and 3 of 12 tumors tested had cooperating mutations in the Ras/MAPK pathway. The residual Dnmt3a allele was neither mutated nor downregulated in these tumors. The bone marrow cells of Dnmt3a+/- mice had a subtle but statistically significant DNA hypomethylation phenotype that was not associated with gene dysregulation. These data demonstrate that haploinsufficiency for Dnmt3a alters hematopoiesis and predisposes mice (and probably humans) to myeloid malignancies by a mechanism that is not yet clear.