Microsoft’s Majorana 2 Chip Achieves 20-Second Parity Lifetime — But the Topological Qubit Debate Rages On

On June 2, 2026, at its Build conference in San Francisco, Microsoft unveiled Majorana 2, its second-generation topological quantum chip. The company claims a 1,000-fold improvement in qubit stability over the first-generation Majorana 1 chip announced just over 15 months earlier, and has accelerated its timeline for a scalable quantum computer from 2033 to 2029.

The bottom line: the materials engineering achievement is real and measurable. Whether Microsoft has actually built a topological qubit, or is instead measuring something else entirely, remains one of the most contentious questions in condensed-matter physics. Several prominent physicists responded to the announcement within hours, saying the new data does not resolve their fundamental objections.

The News

June 2, 2026 – Microsoft Quantum released a technical paper reporting Z-parity lifetimes exceeding 20 seconds in an InAs–Pb tetron device, an improvement of more than three orders of magnitude over the 1–12 milliseconds measured in the aluminum-based Majorana 1 devices reported in previous papers.

The central hardware change is a swap of the superconducting material from aluminum to lead. Lead’s parent superconducting gap (~1,300 µeV) is roughly four times that of aluminum (~300 µeV), and the paper reports that this translates into a doubled topological gap: approximately 70 µeV in the top quintile for the lead-based devices, compared with approximately 30 µeV for the prior aluminum-based platform.

The paper also introduces a GaSb substrate replacing InP, a composite quantum well (6 nm InAs + 2 nm InAs₀.₈Sb₀.₂), and an rf-based tuning technique for characterizing low-energy wire states. The fabricated layout is a prototype four-tetron unit cell for a larger array, with neighboring tetrons coupled through shared quantum dots. The experimental results reported in the paper, however, are narrower: the rf spectroscopy and Z-parity lifetime measurement were performed on a single nanowire of one grounded tetron (BA), not on the full unit cell operating as a multi-qubit processor.

Microsoft announced that based on this progress, it has cut its timeline in half for delivering a scalable quantum computer, now targeting 2029. Microsoft is one of two companies advanced to the final phase of DARPA’s Underexplored Systems for Utility-Scale Quantum Computing (US2QC) program.

My Analysis

I have covered Microsoft’s topological quantum program extensively over the years, including detailed analysis of the Majorana 1 data and a comprehensive assessment in my topological quantum computing modality article. My position has become increasingly skeptical of Microsoft’s claims, and the Majorana 2 paper does not change that. But I want to be clear about what I am skeptical of and what I am not, because the paper contains interesting materials science work that deserves careful examination on its own terms.

What the Paper Actually Shows

The paper makes three core experimental claims, and all three appear well-supported by the presented data.

First, the materials improvement is substantial. The switch from aluminum to lead as the superconductor, combined with the GaSb substrate and the composite InAs/InAsSb quantum well, produces measurably better device properties. The parent superconducting gap of their epitaxial Pb films is approximately 1,300 µeV, compared with approximately 300 µeV for aluminum. The proximity-induced gap in the semiconductor reaches approximately 570 µeV in the nanowire geometry at zero magnetic field. Buried quantum well Hall bars show electron mobilities exceeding 350,000 cm²/Vs. Localization lengths in the lowest occupied subband exceed 1 µm, comparable to the full device dimensions and well above the minimum topological coherence length expected for a clean system (~100 nm).

These are strong materials numbers. The team reports a topological gap of approximately 70 µeV (top quintile) in the Pb-based devices, more than double the approximately 30 µeV they measured in their aluminum-based platform. The region passing their Topological Gap Protocol (TGP) exceeds 1.1 mV·T in the (V_p, B) parameter space, again more than double the aluminum-platform result.

Second, the parity lifetime improvement is dramatic. In the aluminum-based devices described in their earlier papers, Z-parity lifetimes ranged from 1 to 12 milliseconds. In the Pb-based device, they measure a characteristic parity switching time of τ_Z = 22 ± 1 seconds, with some individual dwell times reaching minute-scale. This three-orders-of-magnitude improvement was extracted from 324 total dwell intervals, fitted to a single-exponential distribution consistent with a homogeneous Poisson process from high-energy quasiparticle poisoning.

The paper attributes this improvement to lead’s larger superconducting gap making Cooper pair breaking more difficult, combined with stronger electron-phonon coupling in lead that accelerates quasiparticle recombination. Both effects reduce the non-equilibrium quasiparticle density, which was the dominant source of parity errors in the aluminum devices.

Third, the rf-based spectroscopy technique for characterizing wire-end states is a practical advance. The method uses one quantum dot as a local probe while a second quantum dot at the opposite wire end acts as a parity injector, allowing controlled switching of the wire parity and measurement of the Majorana splitting energy E_M with approximately 1 µeV resolution. This is considerably better than the approximately 7.6 µeV resolution achievable with temperature-broadened DC conductance measurements at 50 mK. The technique is compatible with scalable, dispersive gate-sensing hardware and is amenable to automation, which matters for bringing up larger arrays.

The paper also describes extended parameter regimes where E_M is below their approximately 1 µeV measurement resolution, consistent with their design simulations.

What the Paper Does Not Show

The more important question is what the paper does not demonstrate, and here the list is considerable.

No X measurements. A functioning qubit requires two complementary types of measurement: Z (parity of a single wire) and X (joint parity involving both wires of the tetron). The Majorana 2 paper presents only Z measurements. Microsoft later reported both X and Z measurements on the Majorana 1 chip in a separate arXiv preprint posted in July 2025, four months after the APS March Meeting where the Majorana 1 Nature paper (which contained only Z data) drew significant skepticism. The X measurement data in that later preprint continued to draw criticism from physicists including Frolov and Legg. The Majorana 2 paper sidesteps this question entirely and acknowledges that “investigating these effects will be an important direction for future work.”

This is a significant omission. Without X measurements, the paper demonstrates a long-lived parity state in a superconducting wire, but does not demonstrate a qubit.

No demonstration that the observed states are topological. The paper’s measurements are consistent with Majorana zero modes in the topological phase, but they are also potentially consistent with trivial Andreev bound states that mimic some signatures of Majoranas. The paper itself acknowledges this ambiguity: its model Hamiltonian (Eq. 1) is “sufficiently general to describe a single low energy state which may be present as part of a trivial or topological phase.” The topological interpretation rests on the combination of TGP results, stable zero-bias peaks at opposite wire ends, and the parameter space mapping, but critics have long argued these signatures are insufficient.

No quantum gates, no entanglement, no algorithms. The paper describes a characterization experiment, not a computation. According to a Science report, Nayak has said the team has unpublished data showing they can control their qubits and run quantum algorithms on their chip. Until those data are published and scrutinized, they are claims, not evidence.

No peer review. The paper was posted on Microsoft’s own website and on arXiv. The Majorana 1 Nature paper went through peer review, and the Nature reviewers explicitly stated that the published measurements “do not represent evidence for the presence of Majorana zero modes in the reported devices.” The paper may be under journal review, but that process is not yet public. To be fair, preprints are standard in physics and many important results appear first on arXiv. But given the history of this particular program, the absence of a peer-reviewed version at launch invites extra scrutiny.

The Skepticism, and Why It Runs So Deep

To understand the reception of Majorana 2, you need to understand the history.

Microsoft’s topological quantum program has been running for nearly two decades. In 2018, a Microsoft-funded team at TU Delft published a paper in Nature claiming evidence of quantized Majorana conductance. In 2021, that paper was retracted after external investigators (Sergey Frolov and Vincent Mourik) demonstrated that data had been selectively presented. An independent review by TU Delft’s Scientific Integrity Committee found that the data had been “unnecessarily corrected” and that the researchers had been “caught up in the enthusiasm of the moment.” A second Majorana paper from the same group was retracted from Nature in 2022.

When Microsoft announced Majorana 1 in February 2025, the company claimed to have built the world’s first topological qubit. The accompanying Nature paper, however, fell short of that claim. As I analyzed at the time, the Nature reviewers explicitly noted that the results did not constitute evidence for Majorana zero modes. When Chetan Nayak presented additional data at the APS March Meeting in March 2025, physicist Sergey Frolov of the University of Pittsburgh posted a detailed rebuttal on BlueSky, calling the data “just noise.” Physicist Henry Legg of the University of St Andrews raised similar objections. My own topological quantum computing modality article concluded: “The peer-reviewed evidence does not demonstrate a topological qubit.”

The Majorana 2 announcement landed in this same contested territory. Within hours:

Henry Legg told Science News: “Nothing in this preprint resolves the fundamental issues.” He added: “Nothing in the presented data proves the existence of a topological qubit or Majoranas in these devices.”

Sergey Frolov told Scientific American: “This new preprint is not based on a research track record that can be considered a solid foundation. When Microsoft is mentioned these days, physicists and quantum computing specialists just chuckle or raise their eyebrows.”

On the other side, Kartiek Agarwal of Argonne National Laboratory told Science News that the new rf spectroscopy technique probing nonlocal properties of the wire-end states was “fantastic progress” and supports the case that these are Majorana zero modes.

Sankar Das Sarma of the University of Maryland, who was persuaded by the Majorana 1 Z-measurement data but acknowledged that “people of goodwill could disagree,” has not yet commented on Majorana 2.

Technical Questions Worth Asking

Setting aside the question of whether the states are topological (which I am not qualified to adjudicate), several technical questions in the paper deserve attention.

The 22-second lifetime is a Z-parity measurement in a grounded tetron. The paper notes that parity lifetimes in floating (charging-energy protected) tetrons “could be even longer, since their smaller superconducting islands may further suppress residual quasiparticle effects.” This is plausible but speculative; the measurement was taken on one of two grounded tetrons in the device (BA).

The parity switching time is seven orders of magnitude longer than typical qubit operation times (~1 µs). If taken at face value, this means millions of operations could occur within a single parity lifetime. But this ratio assumes the qubit operations themselves do not accelerate quasiparticle generation, which would need to be verified during actual gate operations rather than passive monitoring.

The scalable unit cell design is preliminary. The paper describes a four-qubit prototype containing four tetrons (two grounded, two floating) interconnected through shared quantum dots. The design is intended to tile into larger arrays, and the paper mentions a 12-qubit array as an example. But the measurements in this paper were performed on a single wire of a single tetron, not on the full four-qubit device operating as a quantum processor.

The τ_X question remains open. The paper discusses the expected relationship between Majorana splitting E_M and the X-parity switching time τ_X. Oddly, the preprint states that τ_X “scales as E_M²” but then says in the next sentence that suppressing E_M “dramatically extends τ_X.” These two statements are contradictory: if the lifetime scaled as E_M², making E_M smaller would shorten it, not extend it. The correct physical relationship is that the error rate scales as E_M², meaning the lifetime scales inversely. This is exactly the kind of inconsistency that peer review is supposed to catch. Setting the typo aside, the paper does not measure τ_X at all, acknowledging that “investigating these effects will be an important direction for future work.” Given that the X measurement was precisely the point of contention with Majorana 1, this omission is conspicuous.

The Topological Gap Protocol is itself contested. A March 2025 Nature article discussed analysis that “pokes holes” in the TGP methodology. The TGP is central to how Microsoft identifies the topological phase in its devices, so challenges to the protocol itself undercut the foundation of the topological interpretation. The Majorana 2 paper acknowledges that its rf tuning protocol is “distinct from and complementary to” the TGP, which it uses for large-scale phase-diagram mapping, while the rf method identifies regions of small E_M for qubit operations.

Reproducibility across devices is not shown. The 22-second τ_Z comes from 324 dwell intervals across repeated time traces on one wire of one tetron. The paper does not show the same lifetime across an ensemble of nominally identical tetrons or across independently fabricated chips. That matters because a central controversy in Majorana nanowires has always been whether interesting signatures are reproducible and robust, or whether they appear only at carefully tuned operating points in individual devices.

The DARPA Dimension

Microsoft’s continued involvement in DARPA’s US2QC program adds a serious engineering-scrutiny signal, though not an independent physics validation of the topological qubit claim. Microsoft is one of two companies advanced to the program’s final phase, following evaluation of architectural designs and engineering plans by experts from the Air Force Research Laboratory, Johns Hopkins Applied Physics Laboratory, Los Alamos National Laboratory, Oak Ridge National Laboratory, Lawrence Berkeley National Laboratory, and Lawrence Livermore National Laboratory. The program is now part of DARPA’s broader Quantum Benchmarking Initiative (QBI).

DARPA’s assessment does not mean the topological qubit has been independently verified. The program evaluates whether an approach can “plausibly” build a utility-scale quantum computer, and the final phase involves building a prototype. But the involvement of these national laboratories and the rigor of the DARPA evaluation process means Microsoft’s approach has received more independent technical scrutiny than most quantum hardware programs. That scrutiny has not disqualified the approach, and that means something.

What This Means for the CRQC Timeline: Nothing

I want to be direct about the implications of this announcement for the question PostQuantum.com readers care most about: the timeline to a cryptographically relevant quantum computer (CRQC).

As I assessed in my topological quantum computing modality article and in the CRQC Quantum Capability Framework: topological qubits are not a factor in any near-term CRQC timeline assessment. Every other modality in my framework has crossed the threshold from theory to experimental demonstration. Topological has not, at least not to the satisfaction of the broader condensed-matter physics community.

Even if we take Microsoft’s claims at face value, the Majorana 2 device is a prototype that has demonstrated a long parity lifetime in a single wire. The path from here to a fault-tolerant quantum computer capable of running Shor’s algorithm at cryptographic scale passes through numerous milestones that have not been reached: two-qubit operations, quantum error correction, logical qubits, and scaling to thousands of physical qubits, each of which requires solving problems that other modalities have been working on for years.

Microsoft’s 2029 target for a “scalable quantum computer” refers to a machine that would be commercially valuable for problems like materials simulation and drug discovery. This is a different and much lower bar than a CRQC capable of breaking RSA-2048. Published resource estimates for that task vary sharply with assumptions: the well-known Gidney–Ekerå 2021 estimate put RSA-2048 at roughly 20 million noisy qubits for an eight-hour attack, while a 2025 Gidney estimate reduces that to under one million noisy qubits under updated algorithmic assumptions. Either way, Majorana 2 does not provide a comparable resource estimate for Microsoft’s architecture, and the gap between a single-wire parity measurement and a CRQC remains vast.

The urgency to migrate to post-quantum cryptography remains unchanged. As I have argued repeatedly, the reason to act now is not any specific Q-Day prediction but the regulatory, insurance, and contractual deadlines that are already set. The Majorana 2 announcement does not alter that calculus in either direction.

Where I Stand

I have spoken with a number of condensed-matter physicists over the past year, and the skepticism about Microsoft’s topological claims is widespread and deeply held. This is not a fringe position. The skeptics include people who were among the first to propose Majorana-based qubits theoretically, and who would have the most to gain from the approach succeeding.

At the same time, I find myself in an uncomfortable position. I am skeptical, but I am not a condensed-matter physicist. I cannot independently evaluate whether the signatures in this paper are consistent with Majorana zero modes or with trivial Andreev bound states. What I can evaluate is the gap between the published evidence and the marketing claims, and on that axis, the pattern continues: the press materials claim breakthrough progress toward a scalable quantum computer, while the paper itself describes a characterization experiment on a single wire of a prototype device that has not been peer-reviewed.

The materials engineering in this paper looks like a real achievement. Replacing aluminum with lead in a superconductor-semiconductor heterostructure and achieving a 1,000-fold improvement in parity lifetime is a notable result regardless of its topological interpretation. If the topological interpretation holds, it validates a core prediction of the theory and puts Microsoft’s approach on a much stronger footing. If it does not, the materials work is still useful science.

I want Microsoft’s program to succeed. The theoretical promise of topological qubits is extraordinary: hardware-level error protection, small qubit footprint, digital control. If it works, it could leapfrog every other approach. After two decades and several false starts, I will need to see peer-reviewed two-qubit operations, independent replication, and X-measurement data that convinces the skeptics before I update my assessment. Until then, Majorana 2 is an interesting materials paper from a team with a complicated history, working on a technology that the broader community has not accepted as demonstrated.

As I noted in my fault-tolerant quantum race analysis, the competition in quantum computing is fierce. Google, IBM, Quantinuum, and QuEra are all making steady, peer-reviewed progress toward fault tolerance using approaches that do not depend on settling a debate about an exotic quasiparticle. Microsoft’s bet may yet pay off spectacularly, but the company will need to publish the data that resolves the controversy rather than routing around it with materials improvements.