Résumé : We introduce a quantum extension of dynamic programming, a fundamental computational method for efficiently solving recursive problems using memory. Our innovation lies in showing how to coherently generate unitaries of recursion steps using memorized intermediate quantum states. We find that quantum dynamic programming yields an exponential reduction in circuit depth for a large class of fixed-point quantum recursions, including a known recursive variant of the Grover's search. Additionally, we apply quantum dynamic programming to a recently proposed double-bracket quantum algorithm for diagonalization to obtain a new protocol for obliviously preparing a quantum state in its Schmidt basis, providing a potential pathway for revealing entanglement structures of unknown quantum states.