DARPA's HARQ Program Just Funded 19 Teams to Build the Quantum Computer Nobody Expected

Every major quantum computing effort in the world is built around a single bet: pick one qubit technology and scale it until the physics breaks or the engineering catches up. Superconducting circuits at IBM and Google. Trapped ions at IonQ and Quantinuum. Neutral atoms at QuEra. Photonics at PsiQuantum. Billions of dollars, one architecture per company, and a tacit assumption that whoever picks the right qubit wins.

DARPA just placed a very different wager. In April 2026, the agency's Microsystems Technology Office launched the Heterogeneous Architectures for Quantum program — HARQ — funding 19 teams from 15 organizations to prove that no single qubit type will ever be enough. Instead, HARQ is building the architectural foundations for quantum systems that combine multiple qubit modalities in a single machine, each doing what it does best, wired together through high-fidelity interconnects.

If the classical computing analogy holds — and it's the one DARPA is explicitly invoking — this is the moment quantum computing gets its CPU-plus-GPU moment.

Why One Qubit Is Not Enough

The homogeneous approach to quantum computing has a problem that gets worse at scale. Superconducting qubits are fast but lose coherence quickly. Trapped ions hold quantum states for minutes but gate operations are comparatively slow. Photonic qubits travel well but are hard to entangle deterministically. Neutral atoms offer massive parallelism but struggle with two-qubit gate fidelity.

Each platform has been engineered to compensate for its weaknesses, but compensation has limits. Error correction overhead — the extra qubits needed to protect against decoherence and gate errors — varies dramatically across architectures. A problem that requires a million physical qubits on one platform might need only ten thousand on another, if the right operations could be routed to the right hardware.

That "if" is what HARQ is designed to solve. The program's thesis: a heterogeneous quantum system — one that uses different qubit types for processing, memory, and communication — could reduce resource requirements by a factor of 1,000 compared to monolithic architectures. That's not a typo. Three orders of magnitude.

The claim is aggressive, but it mirrors what happened in classical computing. Modern data centers don't run everything on CPUs. They route workloads to GPUs for parallel computation, TPUs for machine learning, FPGAs for custom logic. The specialization made possible by heterogeneous classical architectures is what unlocked the AI revolution. DARPA is asking whether the same principle applies to qubits.

Two Workstreams, One Architecture

HARQ operates through two parallel technical areas over a 24-month performance period, each attacking a different half of the heterogeneous quantum problem.

MOSAIC — Multi-qubit Optimized Software Architecture through Interconnected Compilation — is the software side. Five teams are developing compilers and software frameworks that can take a quantum algorithm and decompose it across multiple qubit types, routing each operation to the hardware best suited for it. Think of it as the quantum equivalent of a heterogeneous scheduler in a modern operating system, but at the circuit level.

The MOSAIC teams include Infleqtion, which secured a $2 million contract to build Multistaq, a cross-modality compiler that extends its existing Superstaq platform. MemQ, Q-CTRL, the University of Michigan, and the University of Pennsylvania round out the workstream. Their goal is to produce "compiled mosaics" — physical circuit layouts that outperform anything a single-platform compiler could generate.

QSB — Quantum Shared Backbone — tackles the hardware challenge. Ten teams are engineering the interconnects that allow different qubit types to communicate without destroying quantum information. This is arguably the harder problem. Quantum states are fragile. Converting a quantum state from one physical representation to another — say, from a superconducting microwave photon to a trapped ion's internal state — introduces noise at every step.

The QSB roster reads like a who's who of quantum physics: IonQ, Harvard, Stanford, UC Berkeley, Carnegie Mellon, the University of Illinois Urbana-Champaign, the University of Texas at Austin, the University of Maryland, the Australian National University, and EPFL in Switzerland. Together, they're developing the transducers, photonic links, and entanglement distribution protocols needed to create a shared quantum communication backbone.

A third component — a government-led architecture study — sits above both workstreams, guiding hardware-software co-design, assessing scalability, and estimating economic impacts.

What This Means for Grant Seekers

If you're a researcher or small business in the quantum technology space, HARQ represents both a direct funding opportunity and a signal about where defense-adjacent quantum money is heading.

Direct participation is still possible. HARQ was solicited through three separate vehicles on SAM.gov: DARPA-PS-25-31, DARPA-SN-25-99, and DARPA-SN-25-98. While the initial 19 teams have been selected, DARPA programs frequently add performers through contract modifications, particularly when the architecture study identifies capability gaps. Program manager Justin Cohen, operating through the Microsystems Technology Office, is the point of contact.

The interconnect problem is wide open. QSB's focus on quantum transduction — converting quantum information between different physical carriers — is one of the hardest unsolved problems in quantum engineering. Researchers working on microwave-to-optical conversion, ion-photon interfaces, or superconducting-to-neutral-atom coupling are working on exactly what HARQ needs. If you have relevant results, this is the time to make them visible to DARPA.

The compiler opportunity is underappreciated. Quantum software has historically been an afterthought compared to hardware. MOSAIC's explicit focus on cross-modality compilation opens a lane for computer scientists, compiler engineers, and quantum algorithm researchers who have been shut out of hardware-centric programs. The $2 million Infleqtion contract suggests individual performer awards in the low single-digit millions — accessible scale for university groups and startups.

Adjacent programs compound the opportunity. HARQ doesn't exist in isolation. DARPA's Quantum Benchmarking Initiative (QBI) is running in parallel, evaluating whether any quantum architecture can achieve utility-scale operation. The QBI expanded with a new Stage A call in March 2026. Researchers whose work touches both heterogeneous architectures and benchmarking have a natural dual-submission path.

The Bigger Strategic Picture

HARQ arrives at a pivotal moment in quantum computing's trajectory. After years of steady hardware improvements, the field is confronting an uncomfortable reality: the path from hundreds of qubits to millions remains unclear for every single platform. Error correction thresholds, manufacturing yield, cryogenic cooling costs, laser complexity — each architecture has its own scaling wall.

The heterogeneous approach sidesteps this by reframing the question. Instead of asking "which qubit type will scale to a million?", HARQ asks "what if we only need each type to do what it's already good at?" Superconducting circuits for fast gate operations. Trapped ions for long-lived quantum memory. Photonic channels for communication between nodes. Each component operates in its sweet spot, and the interconnect fabric handles the rest.

This isn't just a technical convenience — it's an economic argument. Building three specialized subsystems, each operating at moderate scale, could be dramatically cheaper than pushing any single technology to extreme scale. The 1,000x resource reduction that DARPA projects would fundamentally change the economics of fault-tolerant quantum computing.

For the defense and intelligence community, the implications are concrete. Quantum computing's most anticipated applications — breaking public-key cryptography, simulating complex molecular systems for drug and materials discovery, optimizing logistics at unprecedented scale — all require fault-tolerant machines that don't yet exist. If HARQ's architectural thesis is correct, heterogeneous systems could reach utility-scale operation years before any monolithic approach.

What Happens Next

The 19 HARQ teams have 24 months to deliver architectural principles, working compilers, and functional interconnect prototypes. DARPA's track record with quantum programs suggests that successful performers will see follow-on funding, potentially through Phase II contracts or transitions to other defense agencies.

For the broader quantum ecosystem, HARQ may prove to be a more consequential program than its modest per-team funding suggests. If heterogeneous quantum computing works, it doesn't just add another approach to the field — it restructures the competitive landscape entirely. Companies that have specialized in a single qubit type become potential subsystem suppliers rather than platform competitors. The value shifts from who has the best qubit to who can integrate the best system.

The 24-month clock started in April 2026. For researchers and startups looking to position themselves, the window for related DARPA proposals, SBIR topics, and collaborative opportunities is open now — tools like Granted can help you identify the quantum computing funding opportunities that align with your specific technical capabilities before the landscape solidifies.