Background Fusion genes form as a result of abnormal chromosomal rearrangements linking previously separate genes into one transcript. The FGFR3-TACC3 fusion gene (F3-T3) has been shown to drive gliomagenesis in glioblastoma (GBM), a cancer that is notoriously resistant to therapy. However, successful targeting of F3-T3 via small molecular inhibitors has not revealed robust therapeutic responses, and specific targeting of F3-T3 has not been achieved heretofore. Here we demonstrate that depleting F3-T3 using custom siRNA to the fusion breakpoint junction results in successful inhibition of F3-T3+ GBMs, and that exosomes can successfully deliver these siRNAs. Methods We engineered 10 unique siRNAs (iF3T3) that specifically spanned the most common F3-T3 breakpoint with varying degrees of overlap, and assayed depletion by qPCR and immunoblotting. Cell viability assays were performed. Mesenchymal stem cell-derived exosomes (UC-MSC) were electroporated with iF3T3, added to cells, and F3-T3 depletion measured by qPCR. Results We verified that depleting F3-T3 using shRNA to FGFR3 resulted in decreased cell viability and improved survival in glioma-bearing mice. We then demonstrated that 7/10 iF3T3 depleted F3-T3, and importantly, did not affect levels of wild-type (WT) FGFR3 or TACC3. iF3T3 decreased cell viability in both F3T3+ GBM and bladder cancer cell lines. UC-MSC exosomes successfully delivered iF3T3 in vitro, resulting in F3-T3 depletion. Conclusion Targeting F3-T3 using siRNAs specific to the fusion breakpoint is capable of killing F3T3+ cancers without toxicity related to inhibition of WT FGFR3 or TACC3, and UC-MSC exosomes may be a plausible vehicle to deliver iF3T3.