DARPA Moves IBM, IonQ, Atom Computing and Diraq Into Stage C of the Quantum Benchmarking Initiative on 7 October 2026: What Government Testing of Their Hardware Must Show by 2033
In December 1907 the U.S. Army Signal Corps issued Specification No. 486 for a heavier-than-air flying machine. It had to carry two people, fly at least 40 miles an hour and stay aloft for an hour, and the price would rise or fall with the speed actually measured. The Wright brothers brought their machine to Fort Myer, Virginia, in 1908. The first trial series ended in September with a crash that killed Lieutenant Thomas Selfridge. In July 1909 a rebuilt machine passed, averaging just over 42 miles an hour, and the Army paid $30,000, including a bonus for the extra speed. The specification turned a claim about flight into something a government team could measure on a field.
The Defense Advanced Research Projects Agency is now trying to do something similar for quantum computing. On 7 October 2026 it announced that four more teams have entered the final stage of its Quantum Benchmarking Initiative, known as QBI. The announcement is short, and the program page behind it is shorter, but together they set out one of the most concrete public tests so far for the claim that a useful fault-tolerant quantum computer is a few years away. This post reads those two DARPA documents closely, with IBM's own release as a third.
Practical takeaway. Six companies, using six different qubit technologies, are now in the stage where a government team tests their hardware against their own designs. Any of them can be awarded up to $300 million in this stage, and DARPA says it will share what it finds with other U.S. government stakeholders. If those findings reach a wider audience, organizations that buy quantum computing time or hardware on the strength of a vendor's roadmap could gain an outside reference point that did not exist before.
What DARPA announced on 7 October 2026: four companies join Microsoft and PsiQuantum in Stage C
The four new Stage C performers are Atom Computing of Boulder, Colorado, which builds scalable arrays of neutral atoms; Diraq of Sydney, Australia, with silicon CMOS spin qubits; IBM of Yorktown Heights, New York, with modular superconducting processors; and IonQ of College Park, Maryland, with trapped ions. They join Microsoft and PsiQuantum, which entered Stage C after taking part in a pilot that preceded QBI, the Underexplored Systems for Utility-Scale Quantum Computing program. Microsoft is pursuing a compact superconducting topological qubit architecture and PsiQuantum silicon-based photonics, so the six Stage C teams cover six distinct physical approaches.
The four promotions follow a Stage B that DARPA describes as yearlong. In it, the agency examined each company's research and development plan, the technical risks that plan has to overcome, how the company proposed to reduce them and which prototypes it would build to show progress. QBI's managing director, Micah Stoutimore, summed up the shift: "Stage C is where we begin moving from plans to proof."
How Stage C differs from Stages A and B: hardware testing by a government team, with awards of up to $300 million
The QBI program page sets out the three stages and their funding. In Stage A a company describes a utility-scale quantum computer concept with a plausible path to realization in the near term, for up to $1 million. In Stage B it writes the research and development plan, names the risks and the prototypes needed to retire them, for up to $15 million. In Stage C it works with the government to verify and validate that the concept can be constructed as designed and operated as intended, and it can be awarded up to $300 million. The ceiling is per performer and is an upper limit; the documents read for this post do not give the amount any company has received.
The verification work is done by an independent team. DARPA runs a separate solicitation for independent verification and validation, open until 30 December 2026, and IBM's release describes Stage C as the phase in which independent experts test and measure quantum hardware, systems engineering and architectures. Stoutimore said his team will scrutinize "the hardware and systems as they develop to determine whether the technology performs the way the models and roadmaps say it should." The program page adds that QBI will communicate its results to other U.S. government stakeholders, which makes the stage a source of information for agencies well beyond DARPA.
Which companies remain in Stages A and B, and how to read the lists
The program page, current as of 8 October 2026, lists Google Quantum AI and Quandela in Stage A, and Nord Quantique, Photonic, Quantinuum, Quantum Motion, QuEra Computing, Silicon Quantum Computing and Xanadu in Stage B. Quandela joined Stage A in May 2026, and a new Stage A solicitation stays open until 30 November 2026. DARPA expects further promotions, including more companies moving into Stage C.
These lists invite rankings, and DARPA warns against reading them that way. A footnote on the program page explains that a company that has finished one stage's milestones is not listed until a contract for the next stage is awarded, so a name can be missing between contracts. The agency also states that QBI is designed to evaluate every viable approach on its own merits and does not set out to pick winners. The lists alone cannot distinguish contract timing from differences in technical progress.
What "utility scale" means in DARPA's test: computational value above cost, by 2033
QBI defines success in economic terms. A utility-scale quantum computer is one whose computational value exceeds its cost, and the question is whether any approach can reach that point by 2033. The definition matters because it rules out the easy wins of the past decade. A demonstration that a quantum processor performed a task no classical computer could match does not qualify unless the task is worth what the machine costs to build and run.
DARPA's own expectation has moved. The program page estimates that, on balance, it is more likely than not that at least one company will build an industrially useful quantum computer by 2033. In the 7 October announcement Stoutimore went further: "we increasingly expect that someone will build a utility-scale quantum computer by 2033." He added that DARPA does not yet know which team or which technical approach will get there. That is the honest state of the field, and a government agency with access to the companies' plans has now said so publicly.
Why IBM's 2029 fault-tolerance date now meets an outside test
IBM's release makes the vendor's side of the arrangement plain. Jay Gambetta, director of IBM Research, said Stage C "validates the strength of IBM's path" and repeated that IBM expects to deliver its first fault-tolerant system, Quantum Starling, in 2029. That is a company forecast. What changes on 7 October is that a government team will now test the hardware behind the forecast as it is built, and report what it finds inside the government.
For buyers this is useful long before 2033. Hospitals, research centers and governments are already signing contracts on vendor roadmaps. Quentir's Medicine Monitor looked at IonQ's agreement to supply SDT with a Superion 256 system, linked in Korean reporting to a cancer center, and the Defense Monitor followed how an F-35A offset is paying for Switzerland's first IBM Quantum System Two. Both purchases rest on what the supplier says its machines will become. Stage C does not certify any of those deals, and its findings go first to government stakeholders. It does create a body of independent measurement that procurement officers, auditors and insurers can ask about. The cryptography community has a stake as well: fault-tolerant machines of the kind QBI tests are the class of computer that post-quantum migration plans are written against, although QBI measures economic utility and says nothing about codebreaking.
How Quentir Reads It
The Signal Corps specification of 1907 did not make the Wrights fly; it made their claim testable and their price depend on the result. QBI works the same way for a far harder machine. Its value lies less in the six names on the Stage C list than in the method: a stated economic definition, a dated horizon, staged funding and a government team that measures the hardware against the company's own design. That method could travel. European programs funding quantum hardware, and public buyers in health and defense, could ask their suppliers for the same thing: a written design, the risks it must retire, and a third party allowed to check.
The open question is how much of Stage C's findings will become public. DARPA promises to share results with other U.S. government stakeholders, and its announcements so far have been short. Quentir will follow the next promotions and any published assessments across the quantum computing, defense and medicine threads. Readers who want every analysis linked here, and the paid editions behind them, under one subscription can find them in the All-access membership.
Sources: Defense Advanced Research Projects Agency, "Four more teams enter Quantum Benchmarking Initiative's final stage", 7 October 2026; DARPA, Quantum Benchmarking Initiative (QBI) program page, performer list current as of 8 October 2026; IBM, "IBM Advances to Stage C of DARPA Quantum Benchmarking Initiative", PR Newswire, 7 October 2026. The 1908-1909 Fort Myer trials, the 1909 flight speed and the purchase price are described in the Smithsonian National Air and Space Museum's 1909 Wright Military Flyer collection record.
Published intelligence, built to inform your own decisions. Published: October 9, 2026.