Topological qubits and Majorana: where does Microsoft really stand?

TL;DR (status as of September 22, 2026): No public, peer-reviewed dataset yet demonstrates a working topological qubit. Microsoft's Majorana 1 (announced February 19, 2025) was published in Nature 638, 651–655 with an unusual editors' note stating the results "do not represent evidence for the presence of Majorana zero modes"; Majorana 2 (June 2, 2026) reports a 22-second parity lifetime on a single nanowire, measured in the Z basis only, in a preprint that is not yet peer-reviewed. On June 24, 2026, Nature published a formal critique of the underlying diagnostic and Microsoft's reply. Rosetta Quantum reads both sides below, with every claim linked and dated.

What exactly has Microsoft announced?

Two chips and a date. Majorana 1, announced February 19, 2025, was presented as the first quantum processor "powered by a Topological Core": an indium-arsenide/aluminum tetron design with room for 8 qubits on the chip, and, per Microsoft's CEO, "a clear path" to a million-qubit processor (Science News, March 2025). Majorana 2, announced at Build on June 2, 2026, replaces aluminum with lead as the superconductor and reports a Z-parity lifetime of 22±1 seconds — against 1–12 milliseconds for the previous generation, a reported gain of roughly 1,000× (PostQuantum, June 2026). Alongside the chip, Microsoft moved its target for a utility-scale machine forward: 2029, from an earlier 2033. That target will remain a declaration until a machine exists; it is a plan, not a measurement.

Every number in the paragraph above is vendor-reported. That is not an accusation — it is a classification, and it is the one that decides how much weight a number can carry. The topological bet itself is coherent and, if it works, important: a qubit whose protection is built into the hardware would need far fewer physical qubits per logical qubit than the superconducting and trapped-ion platforms Rosetta mapped in the 2026 hardware census.

What does the Nature paper actually show — and what does its editors' note say?

It shows a measurement technique — not a topological qubit, by the paper's own words. The February 2025 paper, "Interferometric single-shot parity measurement in InAs–Al hybrid devices" (Nature 638, 651–655; arXiv:2401.09549), demonstrates single-shot interferometric readout of fermion parity, which is a genuine and necessary ingredient of the tetron architecture. The authors themselves write that the measurements "do not, by themselves, determine whether the low-energy states detected by interferometry are topological" (paper text, via the Majorana 1 review). Nature's editors attached a note stating the results "do not represent evidence for the presence of Majorana zero modes" — a rare editorial caveat on a flagship industry paper.

The distance between that language and the announcement language ("a new state of matter", "quantum's transistor moment") is the core of the dispute. At the APS March Meeting in March 2025, the modal expert reaction to the accompanying data was scepticism: physicists quoted by Science News and APS Physics said the data did not yet distinguish topological Majorana modes from topologically trivial Andreev modes — the exact ambiguity that sank an earlier claim: the 2018 Nature paper on "quantized Majorana conductance" from Microsoft-affiliated researchers was retracted in 2021.

What do the critics dispute?

The diagnostic, not the physics. Nobody in the published exchange disputes that topological qubits are possible in principle; the dispute is whether Microsoft's data shows its devices are in the topological phase. Henry Legg (University of St Andrews) first flagged problems in arXiv:2502.19560 (February 2025), and on June 24, 2026 Nature published his peer-reviewed Matters Arising: the "topological gap protocol" used to select devices contains, per Legg, two coding errors — an antisymmetrization step that indexes bias voltage by array position rather than physical value, and plotting code that displays only the largest candidate region — and the raw public transport data looks "disordered and gapless in appearance" (Nature news coverage, June 2026).

Microsoft's reply, with hardware lead Chetan Nayak as corresponding author, reframes rather than concedes: the protocol was "a tuning procedure", not the evidence; the evidence is a radio-frequency quantum-capacitance signal oscillating with a regular magnetic-flux period. On the code, Microsoft concedes a single off-by-one discrepancy and reports it shifts extracted gap values by less than 5 µeV for more than 96% of pixels (reply, June 24, 2026). The two sides are now arguing from different directions: Microsoft reasons forward from the observed signal ("the signal requires a gap"), Legg insists the gap must be established independently (QuantumZeitgeist's 2026 review of the dispute). Both positions are in the open literature, which is exactly where this should be settled — a point Rosetta Quantum has made before about how credible claims get tested.

What changed with Majorana 2?

The lifetime, the material — and one axis is still missing. The June 2, 2026 preprint reports a Z-parity characteristic lifetime of 22±1 seconds over 324 dwell intervals, with some dwells reaching minute scale, on a single nanowire of one grounded tetron; lead's larger parent superconducting gap (~1,300 µeV vs ~300 µeV for aluminum) supports a reported topological gap of ~70 µeV, up from ~30 µeV in the aluminum devices (PostQuantum's technical readout). Those are real engineering numbers, and if they replicate across devices they will matter.

What the preprint does not contain: X-basis measurements (acknowledged by the authors as "an important direction for future work"), two-qubit operations, or braiding (The Quantum Insider, June 2, 2026). A qubit is a two-axis object; public data so far measures one. The critics' June 2026 reactions track that gap: Legg — "Nothing in this preprint resolves the fundamental issues"; Sergey Frolov — the preprint "is not based on a research track record that can be considered a solid foundation" (PostQuantum, June 2026). As of September 22, 2026, the Majorana 2 paper remains a vendor preprint without peer review.

What would settle the debate?

Four things, all of them public. First, X-basis and Z-basis measurements on the same device — that is the minimum for calling the object a qubit, and Microsoft itself lists it as future work (TQI). Second, peer review of the Majorana 2 results. Third, replication statistics across devices and chips — the current record is one wire of one tetron (PostQuantum). Fourth, the DARPA US2QC evaluation, where Microsoft is one of two companies in the final phase and national-laboratory evaluators examine proprietary data — if that program publishes findings, they will be the closest thing to an independent verdict this architecture has had.

Until then, the honest scoreboard reads as the hero of this post does: engineering progress that is vendor-reported and real-looking, a diagnostic under formal published dispute, and zero peer-reviewed demonstrations of a complete topological qubit — on any platform, by any vendor, as of today. For how that compares with the platforms that do publish qubit demonstrations, see Rosetta's superconducting vs trapped-ion comparison.

What we don't know

Rosetta Quantum has not measured any of this hardware; this post audits the public record, not the devices. We do not know whether the modes in Microsoft's devices are topological — as of September 22, 2026, nobody outside Microsoft's proprietary-data reviewers can know that from public data, and that is precisely what the Nature exchange is about. We do not know the device-to-device statistics behind Majorana 2's single-wire result. We do not know what DARPA's evaluators have seen: more than 50 technical evaluators examine data that is not public (PostQuantum), so their eventual verdict could move this story in either direction. And 2029 is a target in the future tense — it will be graded when it arrives, not before.

Rosetta Quantum publishes verdicts with raw, reproducible data. This is educational content, not a product claim. Sources below; corrections with evidence are welcome.