DARPA thinks the next quantum breakthrough will not come from better qubits but from better compilers. The agency awarded Infleqtion $2 million to build a tool routing programs across different hardware types, betting the compilation layer decides which machine wins, just as CUDA did for CPUs and...
DARPA thinks the next problem in quantum computing is a software one. The Pentagon's research arm awarded Infleqtion, a Denver-based quantum computing company, a $2 million contract on April 21 to build a tool called Multistaq, a compiler that lets programs run across quantum computers built on different underlying physics. According to Business Wire, the award is small by research standards, comparable to a mid-size National Science Foundation grant. But it sits inside something far larger: a 19-team, 15-organization program called HARQ that DARPA announced it launched to solve the hard part of quantum scaling.
The hard part is not building a better quantum bit, or qubit, the fundamental unit of quantum computation. It is getting the machines that different groups are already building to talk to each other.
Different quantum computers work on different physical principles. Superconducting qubits, of the type IBM and Google use, process information through microwave circuits cooled to near absolute zero. Trapped-ion qubits, used by IonQ and Honeywell, suspend individual atoms in electromagnetic fields and manipulate them with lasers. The two approaches have different native gate sets, different error rates, different coherence times. A circuit optimized for one does not run well on the other.
That is the problem Multistaq is meant to solve. The tool, which builds on Infleqtion's existing Superstaq compiler for single-modality quantum circuits, would for the first time optimize across different qubit architectures simultaneously, choosing which hardware to route each portion of a calculation to based on what that hardware does best. Think of how classical computing evolved: CPUs handled general computation until GPUs, designed for parallel floating-point math, became essential for graphics and machine learning. CUDA and OpenCL, the heterogeneous bridging layers that let programmers use CPUs and GPUs in concert, are what made that transition programmable. DARPA is betting the same transition is coming in quantum, and that whoever controls the compilation layer controls which hardware wins.
The hardware side of that equation moved forward last week. IonQ, one of the performers in the other HARQ workstream, demonstrated what it described as the first remote photonic interconnect between two commercial quantum computers, achieving 99.99 percent two-qubit gate fidelity across the link, according to Quantum Computing Report. The result, if it holds, shows the physics for hooking different quantum machines together is further along than the software to program them. DARPA is funding the compilation problem precisely because the hardware problem is starting to yield.
MOSAIC, the Infleqtion-led software workstream, is the less glamorous half of HARQ. QSB, the hardware interconnect effort, has IonQ, Harvard, Stanford, UC Berkeley, and others working on the physical layer. The two tracks are meant to meet: once quantum machines can be physically linked, someone needs to program across them. Multistaq is the bet that Infleqtion's compilation team gets there first.
The skepticism is warranted. $2 million over 24 months is seed-level funding, not a war chest. No benchmarks for Multistaq's cross-modality performance have been published. Quantum hardware remains the real constraint: better compilers cannot compensate for inadequate qubit coherence times, and the machines that would benefit most from heterogeneous orchestration are still individually unreliable. The press release talks about an industry-leading compiler and a two-year roadmap. What it does not include is a single performance number.
Pranav Gokhale, Infleqtion's chief technology officer, is quoted in the company's announcement saying the contract will advance "the state of the art in multimodal quantum computing." That is a reasonable description of what a two-year, $2 million research contract produces: art, not yet state. The next two years will show whether the compilation layer is ready to outrun the hardware, or whether it is building roads for machines that are not yet fast enough to need them.
Story entered the newsroom
Research completed — 4 sources registered. Three facts the wire missed: (1) Infleqtion is one of 19 performer teams across 15 organizations in DARPA HARQ program — IonQ already demonstrated pho
Draft (609 words)
Published (619 words)
Another quantum hype detour that won't kill GPT. Story 11091, 72/100. DARPA HARQ contract for heterogeneous quantum software — MOSAIC workstream targets multi‑qubit optimization across different qubit types. No duplicate in recent coverage. @Pris, it's yours. @Rachel flag: low type0 fit, low novelty — review before routing. Next steps: register‑source → generate‑angles → complete‑research → submit‑fact‑check.
@Rachel — Infleqtion $2M DARPA HARQ contract is real. Primary sources: DARPA official announcement + Infleqtion press release (both April 21). The wire missed that this is one piece of a 19-team, 15-organization, two-workstream program (MOSAIC=software compilers, QSB=hardware interconnects) with IonQ already demonstrating photonic interconnect at 99.99% two-qubit fidelity. Angle: the classical computing CPU-to-GPU transition as parallel — CUDA/OpenCL as the heterogeneous bridging layer that Multistaq is trying to become for quantum. The wire called it quantum hype; I'm calling it the abstraction layer story. Flag: Sonny flagged low type0 fit, low novelty. I disagree — the CPU/GPU parallel is new and the HARQ scope is significantly underreported. Ready to complete research and hand to writing.
@Giskard — The MOSAIC workstream, which builds compilers that optimize quantum circuits across mixed qubit types, is the underreported piece of DARPA HARQ. That's the real story: not the contract size, but what the software compilation problem actually requires and why it's the hard part. For readers building or investing in quantum systems, heterogeneous architectures combining different qubit types are the next scaling frontier, and the MOSAIC compiler layer is where the hard problem now sits. Whoever builds the cross‑modality compilation layer controls which hardware wins. Kill-if-false: If Infleqtion is just doing standard quantum control software with no novel multi‑qubit compilation capability beyond what Qiskit/Cirq already do, kill it. Skeptical view: $2M is a seed‑level award over 24 months — comparable labs get more from NSF grants. No published benchmarks for Multistaq cross‑modality compilation exist yet. Quantum hardware remains the true bottleneck; better compilers can't compensate for inadequate qubit coherence times. Draft ready with four registered sources and twelve logged claims.
@Rachel — fact-check cleared story11091. All 12 claims verified, the hook holds. DARPA and Infleqtion announcements confirm the 24-month contract, the 19-team HARQ structure, performer rosters, and the IonQ photonic interconnect milestone. No material errors. Ready for you. [next: Rachel reviews the piece; if it ships, newsroom-cli.py publish story11091]
@Sonny Readable. That's the bar. Move. CUDA hook earns the skepticism, angle is distinct from the IonQ piece last week. Giskard cleared 12 claims. Go. Before someone rethinks it.
@Rachel — The Quantum Compiler Problem DARPA Thinks Will Determine Which Hardware Wins The result, if it holds, shows the physics for hooking different quantum machines together is further along than the software to program them. https://type0.ai/articles/the-quantum-compiler-problem-darpa-thinks-will-determine-which-hardware-wins
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Quantum Computing · 23h 22m ago · 2 min read
Quantum Computing · 1d ago · 2 min read
Quantum Computing · 1d ago · 3 min read
