Microsoft has placed a Majorana 2 system in a dedicated DARPA space, separate from its own labs. Neither party says what will be measured or whether results go public.
Microsoft Quantum opened a 15,000-square-foot research center in the University of Maryland's Discovery District on September 22 and, in the same announcement, said it has delivered a topological quantum system built on its Majorana 2 chip for independent test and evaluation by DARPA. According to the post by Charles Tahan, a partner at Microsoft Quantum, "DARPA will have full access to our latest topological system on site."
That sentence is the news: DARPA's evaluation team has worked with Microsoft since early 2023, by DARPA's account, but until now it did so at Microsoft's own sites or remotely. Maryland gives the agency a system of its own to run, in its own space, with a mandate to evaluate. It arrives just over three months after Microsoft pulled its timeline for a scalable machine forward to 2029, while several independent physicists said the company still had not demonstrated that its device contains a topological qubit at all.
What the announcement does not contain matters as much. Microsoft's post names no test protocol, evaluation period, commitment to publish results, qubit count for the delivered system, or dollar value for its agreement with DARPA. A Microsoft executive gave Reuters a fuller description of the arrangement the same day, covered below, but it stops at the same line. DARPA has published nothing about the delivery. Supercomputing News found no statement from either party on what will be measured.
Tahan's post describes the delivery in two paragraphs. Microsoft is providing topological qubits built on its Majorana 2 chip to DARPA's test and evaluation team, which he describes as bringing expertise from hardware and controls through software and applications. The center's floor plan reserves space for DARPA's topological system and for future prototypes, alongside Microsoft's proprietary research areas, partner labs, and a training laboratory. Tahan frames the handoff as "an important step in evaluating our approach toward a large-scale and economically relevant quantum computer."
The post also lists who is involved on the government side. According to Microsoft, the evaluation effort brings together the Air Force Research Laboratory, the Johns Hopkins University Applied Physics Laboratory, and national laboratories including Los Alamos, Oak Ridge, Lawrence Berkeley, and Lawrence Livermore. That list comes from Microsoft. DARPA's own releases on the program, from December 2023 and February 2025, name no specific laboratories; the February 2025 release describes a government team of more than 50 experts. Microsoft's own lists have also shifted. Its February 2025 Majorana 1 post named NASA Ames Research Center among the program's participants and did not mention Lawrence Berkeley or Lawrence Livermore. Neither party has stated whether the six institutions Microsoft names are contracted verification performers or collaborators.
Two things the delivered system is not, on the available record. It is not described as the fault-tolerant prototype Microsoft committed to build in the program's final phase; the September 22 post calls it "our latest topological system," and Tahan told Constellation Research the system is for testing and evaluation and could grow over time to larger prototypes. The February 2025 Majorana 1 announcement framed that prototype as a goal for the phase, to be built within years. And it is not described by qubit count. Majorana 1 was announced as eight topological qubits on a chip designed for one million. Nothing in the September post says how many qubits the Maryland system holds or how they are configured.
The delivery falls under the final phase of DARPA's Underexplored Systems for Utility-Scale Quantum Computing program (US2QC), which Microsoft's post describes as part of the agency's broader Quantum Benchmarking Initiative, announced in July 2024. Microsoft's post says it is one of the first companies to reach that phase. DARPA's February 2025 release is more specific: exactly two companies, Microsoft and PsiQuantum, were selected, and the phase is named "validation and co-design of an industrially useful quantum computer." DARPA's Stage B selection page from November 2025 restates that the final phase of US2QC has the same technical goals as Stage C of QBI, in which a government verification and validation team determines whether a company's utility-scale concept can be built as designed and run as intended.
The word co-design is DARPA's, and it describes a posture that differs from a neutral certification lab. In DARPA's framing, the government team is both a development partner and an evaluator. Joe Altepeter, then the program manager, said in February 2025 that the government evaluation team had been working with both companies "since the beginning of 2023" to scrutinize their approaches. The QBI program page now lists Micah Stoutimore, who DARPA said in March 2026 was taking over as managing director from Altepeter, the founding program manager, ahead of Altepeter's departure from the agency. The same page lists an open solicitation, DARPA-PA-26-02-01, for independent verification and validation work, with a December 30, 2026 deadline. The solicitation's existence suggests DARPA is still expanding its verification bench as the first hardware arrives, though that is an inference from the posting, not a statement from the agency.
Whitney Mason, whom DARPA's staff page lists as director of the Multi X Office, the office the QBI and US2QC program pages sit under, appears in Microsoft's and the University of Maryland's photos of the opening. Governor Wes Moore's press release on the ribbon cutting does not list her among attendees, and it does not mention DARPA testing or Majorana 2 anywhere in its text.
The physics dispute around Majorana 2 is narrow and technical, which is why this handoff is a story, not a real-estate announcement.
Microsoft's Majorana 2 post by Chetan Nayak describes a new material stack, with lead replacing aluminum as the superconductor and an indium arsenide and indium arsenide antimonide semiconductor region, and reports parity lifetimes exceeding 20 seconds and occasionally passing one minute, compared with 1 to 12 milliseconds for Majorana 1. The company says its qubits are 1,000 times more reliable than the prior generation and that the topological gap has more than doubled. The supporting preprint, arXiv:2606.03884, was posted June 2, 2026. It is not peer-reviewed.
A parity lifetime measures how long the device holds a particular quantum property before noise flips it. The preprint reports that measurement in one basis, the Z-parity, on one nanowire device. A qubit, by definition, has to be measurable in two complementary bases, and critics quoted in SCN's June coverage said the missing X-parity measurement is the central gap. Henry Legg of the University of St Andrews said at the time that "nothing in this preprint resolves the fundamental issues," and Sergey Frolov of the University of Pittsburgh said the preprint was not built on a research track record he considered a solid foundation. That record, as SCN's June piece laid it out, includes two Delft Nature papers retracted in 2021 and 2022 and the statement in the peer-review file of the February 2025 Majorana 1 paper that its results do not represent evidence for Majorana zero modes. Kartiek Agarwal of Argonne National Laboratory, in the same coverage, described the readout technique itself as real progress. Readout is where the field has moved fastest; a Delft and CSIC team demonstrated single-shot Majorana parity readout using quantum capacitance in February, with parity lifetimes beyond a millisecond.
The underlying question is whether the parity signal comes from Majorana zero modes, which would be topologically protected, or from a trivial Andreev bound state that can mimic the signature. Long Z-parity lifetimes do not distinguish the two. An X-parity measurement consistent with theory on the same device would move the argument.
This is the point at which the DARPA delivery meets the June dispute, and also where the record stops. What has changed, on the published record, is who can operate the device and where. Reuters, reporting from an interview with Zulfi Alam, corporate vice president of Microsoft Quantum, described the shift as one from remote evaluation of machines in Redmond and Europe to a dedicated DARPA space where the agency's experts can load hardware and run it through their own boot sequences; Alam said the arrangement lets DARPA "kick the tires," that the agency will start with the Majorana 2 chip, and that the system is built to take later chips. A government team that can boot the hardware on its own is structurally capable of running whatever measurements Microsoft can run, including the one critics have asked for. Whether it will, over what period, and whether anyone outside government will see the result are all undisclosed. One outlet has gone further, writing that resolving the topological question would require DARPA to perform X and Z parity measurements on the same device. That is the outlet's analysis; no DARPA or Microsoft source has said what will be measured.
The other company in the US2QC final phase has disclosed more about its arrangement. PsiQuantum announced on July 22, 2026, a $125 million expanded agreement with DARPA, a performance-based award covering testing and evaluation of its utility-scale hardware designs, critical components, system-level performance, and software development, along with investment in its facilities in Milpitas, California, and Chicago. It had announced a $31.8 million Stage C agreement in September 2025. Microsoft has disclosed no dollar value and no milestone structure for its own final-phase agreement in the February 2025 Majorana 1 post, the September 2025 Maryland commitment, or the September 22 post.
That asymmetry is a fact about disclosure. It is not a verdict on either program, and the two companies' architectures, photonic and topological, are different enough that a comparison of dollar figures would say little about technical standing even if both were public.
Stoutimore said in DARPA's March 2026 release, which PsiQuantum's announcement quotes in part, that "it now seems likely that someone will build a utility-scale quantum computer by 2033, but it remains unclear exactly which team or teams might get across that finish line." That is a program-level remark; DARPA's stated QBI question is whether an industrially useful machine can be built by that year. Microsoft's own target is 2029. The four-year gap between the company's timeline and the agency's evaluation horizon is worth holding in mind as the Maryland work proceeds, because nothing published so far says which clock the evaluation runs on.
Read alongside the same day's news that the Department of Energy will pay for up to 100 lab-verified logical qubits by 2028 under its Genesis Q competition, with national laboratory staff running the circuits, the Maryland delivery fits a pattern. Two US agencies are now structuring quantum programs so that a government team, not the company, produces the number that counts. DARPA's HARQ program for mixed-modality qubits runs on the same logic of agency-run assessment. For a field where claims have outrun reproducible evidence more than once, and where the error-correction literature is splitting across code families faster than hardware can test them, independent measurement has become the thing worth reporting.
It is also, in this case, a sovereignty arrangement of a specific shape. According to Microsoft's post, the State of Maryland acquired and renovated the building under Governor Moore's Capital of Quantum Initiative, which the governor's office describes as targeting more than $1 billion in investment, with over $500 million pledged by the state and the university. A public university hosts the space, a federal agency evaluates inside it, and a company owns the hardware. Europe is sorting its own quantum hardware bets by funding instrument rather than by qubit type; the Maryland model sorts by who holds the keys to the lab. The governor's release also references up to $100 million in DARPA matching grants. That figure comes from an April 2025 agreement establishing the Capital Quantum Benchmarking Hub at UMD's Applied Research Laboratory for Intelligence and Security, under which DARPA and Maryland agreed to match contributions of up to $100 million each over four years, contingent on results and with no funding level required. Neither party has said whether any of that money touches the Microsoft system. DARPA's release says the hub is to be based at ARLIS, a university-affiliated research center, and described it at signing as an additional testing facility for its experts to evaluate the plans and systems of companies moving through QBI.
The center's other functions are supporting context. Microsoft is opening a quantum hardware makerspace with initial partners AMD, Bluefors, Intel, IQM, Fermilab, Riverlane, and Quantum Motion, aimed at hands-on training on real hardware. Fermilab's same-day release describes extending its open-source QICK control and readout kit, originally designed for superconducting qubits, across other platforms and into education with UMD; it does not mention topological hardware or DARPA. Microsoft is also launching an annual workshop series on measurement-based quantum computing, covering measurement speed and fidelity, error-correcting codes, and verification and validation protocols, and funding its first quantum postdoctoral positions at UMD. The post's closing sections restate the company's platform strategy, including the Atom Computing neutral-atom machine Magne, planned at 50 logical qubits and slated for QuNorth in Denmark by early 2027.
Microsoft committed in September 2025 to a Maryland center that would house "one of the first prototypes of Microsoft's topological quantum computer." A year later, the building is open, and a system is inside it, which is delivery on schedule as stated. The open items are the ones that would let anyone outside DARPA judge the qubit: a published evaluation scope, a timeline, a result, and an X-parity measurement on a second device from any source. None has a date.
The SCN Staff is a small AI editorial squad working under human direction. Each agent owns one job.
Scout does the research. It runs down primary sources and checks what's already been published, on SCN and everywhere else, before a story gets written. If a claim can't be traced back to a real document, Scout flags it.
Forge writes. It takes what Scout found and turns it into a draft, argument and sentences and all. Every SCN piece starts here, then gets sharpened.
Cipher handles search: the titles, descriptions, and keyphrase work that decides whether a good article ever gets found. Least glamorous job on the squad. Also one that matters more than it looks.
Pixel makes the visuals. Images, charts, the occasional diagram, all built to SCN's brand instead of pulled from a stock library. When something's easier to see than to read, it goes to Pixel.
Editorial judgment and the final call stay with the humans. So does the fact-checking.
