In quantum computing, combining different qubits delivers better performance

The Defense Advanced Research Projects Agency has launched a multiyear push to break one of quantum computing’s most stubborn bottlenecks: the dominance of single-technology machines. The new Heterogeneous Architectures for Quantum (HARQ) program aims to knit together different types of qubits—each chosen for what it does best—into unified systems designed for practical, scalable performance.

“Qubit technologies each have their own distinct advantages, but no single approach can deliver everything needed for large-scale, high-performance quantum systems. HARQ is asking the community to shift away from a ‘one-qubit-to-rule-them-all’ mindset,” said DARPA Program Manager Justin Cohen. “We aim to define what a truly heterogeneous quantum architecture looks like and to develop the interconnects that make those systems possible. If successful, this approach could provide a far more efficient path to scaling quantum computing and unlock applications that remain out of reach today.”

DARPA’s plan mirrors a lesson from classical computing, where heterogeneous systems—CPUs joined by GPUs, FPGAs, and ASICs—deliver speed and efficiency by matching workloads to the hardware built for them. In quantum computing today, most platforms are still homogenous: trapped ions, superconducting circuits, neutral atoms, photonics, or spins, each with strengths but also hard limits that force compromises in system design.

HARQ splits its work into two tracks. The software-focused effort, called Multi-qubit Optimized Software Architecture through Interconnected Compilation (MOSAIC), will build compilers and frameworks that partition and schedule quantum circuits across diverse qubit modalities to cut resource counts and improve performance. The hardware effort, Quantum Shared Backbone (QSB), targets the biggest physical hurdle: creating high-fidelity links that let unlike qubits talk to each other inside one machine.

Nineteen performer teams from 15 organizations will participate across the two streams. The MOSAIC roster includes Infleqtion, MemQ, Q-CTRL, the University of Michigan, and the University of Pennsylvania. QSB teams include Australian National University, Carnegie Mellon University, École Polytechnique Fédérale de Lausanne, Harvard University, IonQ, Stanford University, the University of California, Berkeley, and the University of Illinois Urbana–Champaign. DARPA said 17 of the 19 teams are already under contract, with two still in negotiation.

Over the next 24 months, the performers will engage in intensive co-design, defining architectural principles, toolchains, and the physical and logical interfaces needed to make cross-qubit systems work. The agency says the goal is to demonstrate the feasibility and scalability of heterogeneity, paving the way for larger demos, future infrastructure investments, and machines aimed at materials discovery, chemical simulation, and biomedical problems, with a clear eye to national security use cases.

The bet on heterogeneity speaks to a mounting consensus in the field: no single modality checks every box. Superconducting qubits typically offer fast gates and tight integration with conventional electronics; trapped ions have demonstrated high-fidelity operations; neutral atoms promise large, reconfigurable arrays; photonic approaches are attractive for networking; and spin-based systems tout stability and potential density. The technical challenge is formidable—linking unlike qubits demands low-noise transduction across disparate frequencies and temperatures, tight synchronization, and control stacks that can compile and route algorithms intelligently without erasing gains through overhead.

If MOSAIC can make heterogeneous compilation routine and QSB can deliver reliable, loss-minimizing interconnects, DARPA’s approach could shift quantum hardware roadmaps away from one-size-fits-all designs toward task-optimized ensembles—much as classical computing did a generation ago.