July 22, 2026 — PsiQuantum announced a 125 million USD expanded tasking agreement with DARPA under the Quantum Benchmarking Initiative (QBI) Stage C, the program’s final phase. The performance-based award is the company’s most valuable US government agreement to date, by its own description, and it follows the 31.8 million USD Stage C award announced in September 2025. PsiQuantum remains one of only two companies left in Stage C; the other is Microsoft, per DARPA’s own selection announcement.
Bottom line: the US government is committing 125 million USD to test whether PsiQuantum’s fault-tolerant photonic design works as drawn. Getting this far means the design survived every earlier, cheaper review stage DARPA ran; it does not mean the verification has succeeded.
The path here ran through three successive gates. DARPA selected PsiQuantum in January 2023 for the first stage of the Underexplored Systems for Utility-Scale Quantum Computing program, the QBI predecessor, moved it to the second stage in January 2024, and advanced it to Stage C in February 2025. QBI’s stated mission is to determine whether any architecture can reach utility-scale operation, defined as computational value exceeding cost, by 2033. Stage C gives DARPA’s audit teams direct access to the company’s facilities and testing work, and the evaluation so far has included prototype design reports, real-time component demonstrations, and on-site testing of fabrication and measurement processes, per the company’s release.
Under the expanded tasking, the audit widens: it funds verification of the utility-scale hardware architecture, custom photonic manufacturing processes, and cryogenic systems, plus live checks of fault-tolerant algorithm resource estimates on Construct, PsiQuantum’s fault-tolerant algorithm development platform, according to Quantum Computing Report. It also pays for infrastructure at Milpitas, California and Chicago, Illinois.
Two pieces of context round out the picture. In May, PsiQuantum signed a Letter of Intent with the Department of Commerce for 100 million USD in proposed CHIPS incentives covering barium titanate optical switches, high-temperature single-photon detectors, and photonic packaging, one of nine quantum LOIs totaling about 2 billion USD. On July 15 the company made Victor Peng, interim chief since February, its permanent CEO.
My Analysis
As far as I can tell from tracking these programs worldwide, QBI Stage C is the most rigorous vetting any fault-tolerant roadmap gets from any government. DARPA spent the earlier stages filtering out approaches whose physics or engineering failed independent review, and PsiQuantum passed every filter. I treat that as real signal. Skeptics who dismissed the company’s photonics approach as slide-ware now have to take the argument up with DARPA’s verification teams, who have been inside PsiFactory. Peng, who ran Xilinx and served as AMD’s president before taking this job, called QBI one of the most rigorous government technology evaluations he has seen. He is paid to say that, and a Xilinx and AMD veteran has also seen plenty of government technology programs up close.
The claim-evidence gap stays wide all the same. PsiQuantum has yet to demonstrate a single logical qubit in public. This award buys verification rather than qubits: DARPA is funding component tests, design audits, and resource-estimate checks on Construct. A design that keeps passing component review is credible, and a credible design is still several hard engineering steps from a working cryptographically relevant quantum computer (CRQC). Under my CRQC Quantum Capability Framework, I assess the architecture as comparatively strong on manufacturability and connectivity, with public evidence gaps on demonstrated error correction. This agreement does not close those gaps. It pays DARPA to measure them.
Most useful for readers is the two-company Stage C field. The finalists could hardly differ more: PsiQuantum builds on silicon photonics and fusion-based computation, while Microsoft pursues topological qubits in a superconducting architecture, and neither resembles the transmon and trapped-ion machines that dominate today’s cloud offerings. DARPA is careful to say QBI is not a winner-take-all competition, and the funnel runs in both directions: in March it reopened Stage A to new entrants, with abstracts due July 31 and full proposals due September 30, while a Stage B cohort including Quantinuum and Silicon Quantum Computing works the middle. Read the shape anyway: the agency wants every credible path audited, and it spends serious verification money only where the audits keep passing. Two companies clear that bar today.
On timing, QBI’s managing director, Micah Stoutimore, said in March that “it now seems likely that someone will build a utility-scale quantum computer by 2033,” while noting nobody knows which team crosses first. Utility-scale is an economic bar, computational value exceeding cost, and not a cryptanalytic one; the first machine over it will almost certainly not be running Shor’s algorithm against RSA-2048. I keep those thresholds separate in my Q-Day analysis, and the separation buys planners less comfort than they might hope. When the government’s own benchmarking chief treats useful fault-tolerant hardware by 2033 as likely, and the July 7 White House Quantum Summit pushed the same urgency, a PQC migration that finishes in 2032 is cutting it fine.
My position: I would rather hold PsiQuantum’s side of this agreement than argue against it (a sentence I could not write about most quantum press releases this year), and I am still not moving my Q-Day estimate on the strength of a verification agreement. Watch for the first loss-inclusive logical-qubit demonstration. Contracts are not qubits.
