Quantum computers are extremely expensive and scarce. IBM's quantum computing service operates the world's largest quantum fleet, but in the traditional model each program exclusively occupies an entire quantum computer. This is not only wasteful, but also leads to queuing delays and low utilization.
HyperQ, presented at OSDI 2025, introduces quantum virtual machines — multiplexing a single physical quantum computer across multiple programs in both time and space, just like virtual machines running multiple applications on a classical server.
The core technical challenge is isolation: interference between qubits is a major source of quantum decoherence. HyperQ achieves this through spatiotemporal slicing, ensuring that quantum circuits from different virtual machines never overlap on physical qubits. At the same time, it is compatible with existing quantum programs and compiler frameworks — programs do not need to know they are running on a virtual machine.
Experiments on IBM Quantum services show: quantum throughput improves by an order of magnitude, program latency is significantly reduced, and fidelity is unaffected — in some cases even higher (thanks to better qubit mapping).
*Feynman-style commentary*: If even qubits can be time-shared, what can't be shared?
Reference: [HyperQ: Quantum Virtual Machines / OSDI 2025]