The trapped-ion quantum charge-coupled device (QCCD) proposal1,2 lays out a blueprint for a universal quantum computer that uses mobile ions as qubits. Analogous to a charge-coupled device (CCD) camera, which stores and processes imaging information as movable electrical charges in coupled pixels, a QCCD computer stores quantum information in the internal state of electrically charged ions that are transported between different processing zones using dynamic electric fields. The promise of the QCCD architecture is to maintain the low error rates demonstrated in small trapped-ion experiments3-5 by limiting the quantum interactions to multiple small ion crystals, then physically splitting and rearranging the constituent ions of these crystals into new crystals, where further interactions occur. This approach leverages transport timescales that are fast relative to the coherence times of the qubits, the insensitivity of the qubit states of the ion to the electric fields used for transport, and the low crosstalk afforded by spatially separated crystals. However, engineering a machine capable of executing these operations across multiple interaction zones with low error introduces many difficulties, which have slowed progress in scaling this architecture to larger qubit numbers. Here we use a cryogenic surface trap to integrate all necessary elements of the QCCD architecture-a scalable trap design, parallel interaction zones and fast ion transport-into a programmable trapped-ion quantum computer that has a system performance consistent with the low error rates achieved in the individual ion crystals. We apply this approach to realize a teleported CNOT gate using mid-circuit measurement6, negligible crosstalk error and a quantum volume7 of 26 = 64. These results demonstrate that the QCCD architecture provides a viable path towards high-performance quantum computers.