A long-simmering dispute over Microsoft’s most ambitious quantum computing claim has finally reached the pages of Nature. On June 24, 2026, the journal published a formal critique by physicist Henry Legg, which directly challenges the 2025 paper that introduced Majorana 1—Microsoft’s first supposedly topological qubit processor. The critique, accompanied by a reply from Microsoft’s research team, thrusts the controversy back into the limelight, raising fresh questions about one of the most closely watched technologies in the race to build a useful quantum computer.
For years, Microsoft has pursued a unique approach to quantum computing: topological qubits. Unlike the superconducting or trapped-ion qubits used by IBM and Google, topological qubits are designed to be inherently protected from environmental noise by encoding information in exotic quasiparticles called Majorana zero modes. Theoretically, this would make them far more stable and require less error correction—an enormous practical advantage. The catch? Majorana zero modes had never been conclusively demonstrated in a working device. Microsoft’s 2025 Nature paper seemed to change that.
The paper, authored by researchers from Microsoft Azure Quantum and several academic collaborators, described a device built from semiconductor–superconductor nanowires that allegedly hosted Majorana zero modes at both ends. They reported a characteristic zero-bias conductance peak—a long-accepted signature—and more nuanced topological gap protocols. At a high-profile press event in Redmond, Microsoft unveiled Majorana 1 as the culmination of nearly two decades of research, comparing the achievement to the invention of the transistor. The stock market reacted, competitors issued cautious congratulations, and the quantum community began re-evaluating timelines. But not everyone was convinced.
Henry Legg, a condensed-matter theorist then at the University of Basel, sounded the first public alarm only weeks after the paper’s release. In a series of preprints and conference presentations, he argued that the same zero-bias peaks could arise from trivial disorder or weak antilocalization—phenomena that mimic Majorana signatures without any topological physics. Legg’s analysis zeroed in on the core claim: the topological gap protocol, the very evidence that Microsoft said separated their signal from false positives. According to Legg, the protocol as implemented was insufficient to rule out trivial Andreev bound states, a well-known source of confusion in nanowire experiments. His detailed simulations, later published in 2026 as the Nature critique, suggested that the experimental data were better explained by non-topological mechanisms.
The formal critique, which underwent peer review itself, marks a rare and significant moment. Nature does not lightly publish rebuttals of its own high-profile papers; such critiques are often rejected or buried in comment sections. That Legg’s challenge passed editorial scrutiny indicates the seriousness of the concerns. In the same issue, Microsoft’s team defended their work, providing additional data and reasserting that the topological gap protocol is robust. They acknowledged the possibility of trivial states but argued that their multi-parameter analysis eliminates them with high confidence. The reply also pointed to follow-up experiments—some not yet published—that they say corroborate the topological interpretation.
A Pattern of Deja Vu
The controversy feels eerily familiar to quantum computing veterans. In 2018, a team led by Leo Kouwenhoven, then at Delft University of Technology and working closely with Microsoft, published a paper in Nature claiming the observation of Majorana zero modes. That paper was retracted in 2021 after other researchers, including Legg, identified similar issues. An internal investigation found that the authors had selectively plotted data, and the original raw data did not support the conclusions. It was a humbling moment for Microsoft’s quantum program, which had heavily promoted the 2018 result. The company promised more stringent verification protocols and set out to build a new generation of devices.
The 2025 Majorana 1 paper was supposed to be the vindication. Microsoft had renamed its quantum hardware effort to Azure Quantum, brought in a new leadership team, and spent four years on a fault-tolerant measurement approach. They ditched the simple conductance-peak hunting that had tripped up the 2018 paper and instead relied on the topological gap protocol—a more comprehensive set of measurements. The company’s scientists, including several long-time Station Q veterans, insisted this time was different. The device, they said, was the result of rigorous blind-testing protocols, with independent verification by a separate internal group using a different measurement setup.
Legg’s critique, however, uses the same publicly available data to mount a systematic counterargument. He does not accuse Microsoft of fraud; rather, his work is a technical dissection of the statistical and physical assumptions underpinning the topological gap protocol. For example, the protocol requires fitting multiple conductance traces to a theoretical model that assumes a perfectly clean nanowire. Legg’s modeling shows that even minor disorder—something unavoidable in real devices—can produce fitting outcomes that look topological in the protocol’s decision space. In other words, the protocol is not specific enough. This is not merely an academic quibble. If the Majorana 1 chip does not host topological qubits, then Microsoft’s entire roadmap—built around scaling to a million-qubit machine by the end of the decade—collapses.
Community Reaction and Market Impacts
Since the Nature critique appeared, the reaction from the condensed-matter and quantum computing communities has been swift and divided. Some experimentalists who have long grumbled about overblown Majorana claims see Legg’s publication as a long-overdue course correction. “The topological gap protocol was always a bit of a black box,” said one researcher who wished to remain anonymous. “Peer review is finally catching up to the complexity of these devices.” Others, including several Microsoft collaborators, have rallied to the company’s defense. They argue that Legg’s simulations rely on idealized parameters that may not accurately reflect the actual device, and that no single counter-model can explain all the data. The debate now rests on which model—topological or trivial—offers the more parsimonious explanation, and whether any conceivable set of measurements can resolve it.
The timing is particularly awkward for Microsoft. The company has invested heavily in quantum, not just in hardware but also in building an Azure Quantum cloud platform and a developer ecosystem. Last year, it announced an early access program for hybrid quantum-classical workloads, with the promise that topological qubits would eventually deliver customers a “logical qubit advantage.” Financial analysts have begun revisiting the quantum line item in Microsoft’s R&D budget, which is rumored to exceed $500 million annually. While quantum computing is a small fraction of the company’s overall revenue, the strategic narrative matters. In the cloud wars with Amazon and Google, the perception of leadership in next-generation computing can influence enterprise adoption.
Legg, now an independent researcher, has become a polarizing figure. Critics point to his history of targeting high-profile Majorana claims and suggest he is motivated by a desire to tear down rather than build. Supporters see him as a necessary gadfly in a field where the prize—a revolutionary new computer—can distort scientific rigor. In an interview with a physics podcast earlier this year, Legg said, “I’m not opposed to topological quantum computing. I’m opposed to claiming success before the evidence is in. If we move forward on a false premise, we risk wasting years and billions of dollars.”
Microsoft’s Reply: More Data, Same Conclusion
Microsoft’s official reply in Nature tries to walk a fine line. It does not dismiss Legg’s concerns outright; instead, it acknowledges the possibility of trivial states but argues that the totality of evidence favors a topological interpretation. The reply includes new analyses of the same data, showing that the topological gap protocol performs well even under a wider range of disorder simulations than Legg considered. They also highlight independent electrical and thermal transport measurements—not included in the original paper—that they say are inconsistent with a purely trivial scenario. The reply repeatedly stresses that the Majorana 1 chip was characterized with multiple gates and magnetic field sweeps, and that the signatures persist over a broader parameter range than trivial models predict.
Still, the reply stops short of sharing raw data at the level Legg claims is needed to fully reproduce the analysis. Microsoft cites intellectual property concerns and the need to protect trade secrets. This refusal has drawn criticism. “If you want the community to accept a claim this big, you have to open the kimono,” said a senior scientist at a rival lab. “Topological qubits are supposed to be the cure for noise, but the evidence for their existence is being drowned in noise of a different kind.”
What Happens Next?
Nature’s decision to publish the critique and reply side by side is not the end of the story. It is likely to accelerate three parallel efforts. First, several independent experimental groups—some at academic labs, others at competitors like Google and Intel—are racing to replicate the exact device design. If they can reproduce the zero-bias peaks but fail to see the same topological gap protocol outcomes, Microsoft’s position will weaken. Second, theorists are already developing improved versions of the topological gap protocol that could more rigorously separate topological from trivial signals. A paper posted on arXiv last week proposes using machine learning to classify the raw conductance maps, bypassing the fitting procedure entirely. Third, Microsoft is reportedly planning a Nature Physics submission with data from a second-generation chip, Majorana 2, that supposedly operates at higher magnetic fields where trivial explanations become less plausible.
The drama also underscores a broader challenge in quantum computing: the difficulty of achieving scientific consensus on any major hardware milestone. Last year, Google’s claim of “quantum supremacy” was challenged by IBM, and earlier this year, a Chinese team claimed a photonic advantage that remains under scrutiny. As the field matures, the community is grappling with how to set standards of proof that are transparent, reproducible, and resistant to corporate hype.
For now, the safest statement is that Majorana zero modes in Microsoft’s device remain unconfirmed. The weight of peer-reviewed evidence, as of June 2026, does not conclusively demonstrate topological qubits. That does not mean they don’t exist—only that more work is needed. Microsoft has promised to release an expanded dataset by the end of the year. In the meantime, the Nature critique serves as a case study in how science corrects itself, and a reminder that even the most celebrated breakthroughs are never final until they survive sustained, adversarial examination.