The Quantum Investor's Notebook logo The Quantum Investor's Notebook
← Back

Saturday, July 4, 2026 · Weekly Paper, Week 2

A Gap, or a Crowded Room? The Dispute Beneath Microsoft's Majorana Qubit

I sat down this week meaning to defend Microsoft. The plan was simple: read the paper at the center of a new dispute, then write up why the company's topological-qubit evidence holds. Instead I found a disagreement serious enough that I changed what I set out to say, and a thread that reaches back further than I expected. The paper in focus is not a Microsoft result. It is a Nature Matters Arising, a formal published critique, by Henry Legg, and it questions part of the experimental foundation under Microsoft's topological-qubit claims.1 The charge is not that the device is "not quantum." It is narrower and sharper: that the signal Microsoft read out may not prove the company reached the protected topological regime a Majorana qubit needs. Nature published Microsoft's reply alongside it, so this week I tried to follow both sides honestly, without deciding in advance who is right. It is not a valuation and not financial advice. Do your own research.

The prize: a qubit the hardware protects

The superconducting gap is the moat.

Start with why this matters. Most quantum computers fight errors after the fact, stacking error correctionTechniques that combine many shaky physical qubits into fewer reliable ones, so a long calculation stays correct. on top of fragile physical qubitsAn actual piece of qubit hardware. On its own it is fragile and makes frequent errors.. A topological qubitA theorized qubit that stores information in a special, spread-out quantum state, making it naturally resistant to errors. Microsoft's bet. would build some of that protection into the hardware itself. The idea rests on Majorana zero modesAn exotic, hard-to-create quantum state that Microsoft's topological qubit depends on. Its existence in these devices is still scientifically disputed., or MZMs: special low-energy states that appear in pairs at the two ends of a superconducting nanowireAn extremely thin wire, billionths of a meter wide, used as the body of Microsoft's qubit device.. A bit of information is stored not in one place but in a shared property of the pair, its fermion parityWhether the number of particles in a system is even or odd. It is the quantity Microsoft's qubit reads out., roughly whether the pair holds an even or odd number of electrons. Spread across the whole wire, that property is hard for local noise to read or corrupt. What keeps the two ends from quietly talking, and keeps stray excitations out, is an energy gap around zero: the superconducting gap. The gap is the moat. Hold on to that word, gap: the whole dispute lives inside it.

What Microsoft measured, and what it means

The whole week turned on one distinction: the gap between what Microsoft measured and what that measurementReading a quantum system, which forces it out of its blend of possibilities into a single definite 0 or 1 and ends its quantum behaviour. means. This is the hinge, so it is worth going slowly.

What they measured, in early 2025, is an electrical signal, not "topology."2 They put a tiny sensor, a small quantum dot, right next to the nanowireAn extremely thin wire, billionths of a meter wide, used as the body of Microsoft's qubit device. and watched its quantum capacitance, essentially how eagerly that dot soaks up a little charge as its energy levels shift. That eagerness settles at one of two values depending on the wire's parityWhether the number of particles in a system is even or odd. It is the quantity Microsoft's qubit reads out., whether the wire holds an even or odd number of electrons. So the number they actually read out is just the difference between those two values,

$$ \Delta C_{\mathrm{Q}} = \bigl| \, C_{\mathrm{Q}}(Z=+1) - C_{\mathrm{Q}}(Z=-1) \, \bigr| , $$

where $Z=\pm 1$ labels the two parity states. Threaded by magnetic flux and watched over time, this radio-frequency signal flips cleanly back and forth between the two values, a random telegraph signalNoise that jumps at random between two fixed levels, like a telegraph key clicking on and off. In a qubit it is the fingerprint of one microscopic flaw switching back and forth., and it repeats every $h/2e$ of flux, the natural rhythm for paired electrons. Microsoft reported reading it in a single shot, with an assignment error near one percent.2 That is a real, well-built measurement and a genuine piece of engineering. Hold on to how clean and structured the signal is, because both sides agree it is there. The fight is not about the signal. It is about what the signal proves.

Measuring a clean two-valued signal is not the same as measuring topology.

Measuring a clean two-valued signal is not the same as measuring topology. The flips are the observation. Calling them "Majorana parity flips, protected by a topological gap" is an interpretation laid on top, and Microsoft is careful about this itself. One line in its own introduction stopped me: the measurement, they write, "does not unequivocally distinguish between MZMs in the topological phase and fine-tuned low-energy Andreev bound statesA harmless zero-energy state caused by ordinary disorder in the wire that can mimic a Majorana signal. It is exactly the false positive the protocol is built to reject. in the trivial phase."2 In plain words, an Andreev bound state is the ordinary look-alike: a boring, non-topological state that can sit at the end of a wire and throw off much the same two-valued signature, with no Majorana anywhere in sight. So the signal on its own cannot settle it. Something else has to show the wire was truly in the protected regime. That something is the topological gap protocolMicrosoft's pre-registered pass/fail test for deciding whether a device has truly entered a topological state hosting Majorana modes, built to reject look-alike false signals. Its reliability is still debated..

The filter, and an older doubt

The topological gap protocolMicrosoft's pre-registered pass/fail test for deciding whether a device has truly entered a topological state hosting Majorana modes, built to reject look-alike false signals. Its reliability is still debated., or TGP, was introduced by Microsoft researchers in a 2021 preprint, and its logic is genuinely good.3 Do not trust one pretty signal at one end of a wire. Look at both ends at once, and across a whole landscape of settings: combine how easily current enters each end with how strongly one end is felt at the other, sweep many magnetic fields and gate voltages, and only call a region "topological" if both ends agree and the device shows a gap that closes and reopens the way a real topological transition should. It was built to catch exactly the Andreev look-alikes that fool single-ended measurementsReading a quantum system, which forces it out of its blend of possibilities into a single definite 0 or 1 and ends its quantum behaviour.. In the chain of evidence, the TGP is the filter that certifies the stage on which everything else is performed.

And here is the find that reframed my whole week: this is not the first time the filter itself has been questioned. Before this Nature exchange, the same critic had already published a separate comment, in early 2025, on Microsoft's earlier 2023 study, the one that first reported devices passing the TGP.4 His concern there was not about any single device. It was about the tool. He argued that the TGP's own definitions of "gap" and "topological" did not match between the published description and the analysis code Microsoft released, and that its yes-or-no verdict could flip with choices such as the magnetic-field range or the analysis settings, choices that are about how you look, not about the device in front of you. One more thing stood out to me. As far as I can find, the original TGP protocol was never published as a standalone, peer-reviewed journal paper, even as it quietly became load-bearing for the claims stacked on top of it.

So the new dispute is not a bolt from the blue. It is the second time the same foundation has been called into question, and that is what gives it weight. Strip the headlines away, and a single standing question runs under both critiques: is the filter that certifies "topological" as solid as the results that lean on it?

The critique: an empty room, and footsteps in it

With that history in mind, the new critique goes straight at the raw material under the TGP tune-up: the transport data, the plain conductance measurementsReading a quantum system, which forces it out of its blend of possibilities into a single definite 0 or 1 and ends its quantum behaviour. that Microsoft's paper did not show.1 Legg reconstructs them, and the picture he draws is the heart of the disagreement.

If the gap is an empty room, these are footsteps inside it.

Picture the superconducting gap as a room. In a clean topological wire, the band of energies around zero should be empty and quiet, and that emptiness is the whole protection: with nothing living at those low energies, local noise has nothing to grab, and the two ends cannot quietly talk to each other. In the reconstructed data, Legg finds the room crowded instead, with conductance spread right across the low-energy range where it should be silent. If the gap is an empty room, these are footsteps inside it. The signal is also lopsided from one end of the wire to the other, the kind of asymmetry you would expect from accidental quantum dots, little puddles of trapped charge that form in a disordered wire.

Why does a crowded room matter so much? Two reasons, and they compound. First, if there is no clean gap, the protection the whole qubitThe basic unit of a quantum computer. Like a 'bit' in a normal computer, but instead of being only 0 or 1 it can be 0, 1, or a blend of both at once. idea depends on is simply not there. Second, and this is the sharper point, those crowding low-energy states are the very look-alikes the exercise was meant to rule out. A fine-tuned Andreev bound stateA harmless zero-energy state caused by ordinary disorder in the wire that can mimic a Majorana signal. It is exactly the false positive the protocol is built to reject. in a messy wire can produce a parity-like, two-valued signal that looks a great deal like the real thing. So a crowded gap is not just untidy, it is dangerous: it is precisely the condition under which a trivial, non-topological device can counterfeit the Majorana signature Microsoft is reading.

Hand-drawn two-panel diagram on an energy axis. Left, labeled clean gap: a nanowire with a Majorana zero mode at each end and an empty, quiet band around zero energy. Right, labeled what the data show: the same band crowded with small low-energy states and a line of red footsteps inside it.
Figure 1: A gap, or a crowded room? In a truly topological wire the energy window around zero is empty and quiet (left). In the reconstructed transport data, extra low-energy states crowd that window (right), the footsteps that, the critique argues, are hard to square with a clean, protecting gap.

That is what ties the knot at the center of the dispute. Reading the radio-frequency signal as protected Majorana parityWhether the number of particles in a system is even or odd. It is the quantity Microsoft's qubit reads out. flips already assumes there is a gap holding the trivial states at bay. But the transport data, Legg argues, look like what a gapless, crowded wire would give you. Use the RF signal itself as proof that the gap is there, and you are leaning on the very thing you are trying to establish. (He also flags a small coding error in the TGP processing, which Microsoft concedes; on its own it dents the tidiness of the map more than the core claim.)1

Microsoft's answer

Microsoft's reply, from a large team, does not give up the core, and its argument deserves to be taken seriously.5 Its central move is to pull apart the two things Legg ties together. The TGP, it says, was only a practical tool for finding good operating points, a way to tune the device, and it played no role in reading out the parityWhether the number of particles in a system is even or odd. It is the quantity Microsoft's qubit reads out.. The real evidence is the radio-frequency signal itself, and that signal, Microsoft argues, is not the kind of random local blip a messy wire tosses off by chance.

The strength of the answer is in the structure of the signal, and it is worth taking on its own terms. It is not just "some low-energy state." It is bimodal, cleanly flipping between two values; it is stable over time; and it repeats in magnetic flux with exactly the $h/2e$ period you expect when the charge in motion is a pair of electrons. Microsoft's point is that a gapless, crowded, trivial wire should not be able to hold that pattern. Pile many states near zero energy and their contributions arrive with essentially random phases, which would blur any clean oscillation away; the tidy $h/2e$ rhythm would fall apart rather than persist. The sheer size of the measured capacitance shift, they add, is itself a fingerprint of a well-developed gap, not a gapless smear. And they close on the point that lands hardest: it is one thing to say the room looks crowded, and quite another to build a specific, trivial mechanism that actually reproduces this signal in full, the stable, flux-periodic telegraph and all. Legg reinterprets the data as gapless, they note, but offers no such alternative model.5

My takeaway: this is not a working Majorana qubit, and it does not claim to be. It adds one floor to the building, single-shot parity readout, a real step. The dispute is whether that signal can carry the topological reading on its own.

I find that a fair challenge, and I do not think it can be waved away. But it does not dissolve Legg's point either, because his case is precisely that the foundation used to license the topological reading, the transport data behind the TGP, does not look gapped in the first place. So the two arguments do not quite meet, and the dispute narrows to one honest question, worth stating plainly: can the radio-frequency signal carry the topological interpretation on its own, or does it still depend on a transport-and-TGP foundation that is itself in doubt?

The investor's read

So I put on the investor's hat. Where does this leave Microsoft?

It sends me back to my earlier notebook piece, "The Quantum Bet Microsoft Can Afford to Lose,"6 and that view only hardened. Microsoft can fund a long, uncertain topological program out of what is, for it, pocket change. But as an outside investor I cannot treat the story as de-risked while its central evidence is argued over this way, and while the filter that certifies "topological" carries its own unresolved doubts.

Hand-drawn diagram of a building as a stack of floors. From the bottom: topological gap protocol, marked with a red question mark; topological region; single-shot parity readout, with an arrow labeled this result pointing to it; and a dashed-outline penthouse at the top labeled working qubit, not yet.
Figure 2: A floor, not the penthouse. A working Majorana qubit sits on single-shot parity readout, which sits on a certified topological region, which sits on the topological gap protocol. This result adds the readout floor; the dispute is over whether the foundation beneath it is solid.

To be fair to the other side of the ledger, Microsoft is still moving. In recent preprints, presented as its own not-yet-verified claims, the company reports a parity lifetimeHow long the even/odd state survives before flipping. Microsoft's headline ~20-second number. of about twenty seconds after swapping aluminum for higher-gap lead,7 and single-shot readoutThe step of measuring a qubit to extract its final 0-or-1 answer. of both parityWhether the number of particles in a system is even or odd. It is the quantity Microsoft's qubit reads out. loops of a two-wire "tetronThe specific four-terminal topological qubit device design that Microsoft built." device.8 That same tetron preprint is candid about the distance left: an assignment error of about sixteen percent on one of its two loops,8 far from the high-fidelity gates that would actually demonstrate the Majorana advantage.

When a single paper carries 167 authors, the science can be impressive and the accountability diffuse at once. That is not a charge of bad faith. It is a reason a specific, serious critique deserves a direct technical answer, not a narrower reading of which figures were shown.

Where I land

I will end wearing both hats. As a physicist, I find the dispute clarifying rather than discouraging: the idea of a hardware-protected qubitThe basic unit of a quantum computer. Like a 'bit' in a normal computer, but instead of being only 0 or 1 it can be 0, 1, or a blend of both at once. is still beautiful, and the parity readout is real, careful work. But the two sides do not yet agree on the one thing that matters most, whether the radio-frequency signal can stand on its own as evidence of topology, or whether it still leans on a transport-and-TGP foundation now questioned twice, not once. That is not a settled fact; it is an open argument. As an investor, I do not need Legg to be right on every point. I need the topological regime shown in a way that no longer rests on a contested filter, and real confidence would look like working gates with fidelity numbers that back the promised protection. Until then, the topological gap protocolMicrosoft's pre-registered pass/fail test for deciding whether a device has truly entered a topological state hosting Majorana modes, built to reject look-alike false signals. Its reliability is still debated. stays on my radar, and the Microsoft bet stays what I first called it: important if it works, and affordable if it does not.

Sources & notes

  1. H. F. Legg, "On the robustness of topological gap detection via transport" (Matters Arising), Nature 654, E22-E26 (2026). DOI: 10.1038/s41586-026-10567-8. Source for the critique: the reanalysis of the transport data underlying the topological gap protocol tune-up; the abundance of low-energy states and asymmetric local conductance read as signatures of quantum dots; the conceded coding artifact in the TGP processing; and the argument that using the capacitance signal to establish the gap reverses the hierarchy of evidence.
  2. Microsoft Azure Quantum, "Interferometric single-shot parity measurement in InAs-Al hybrid devices," Nature 638, 651-655 (2025). DOI: 10.1038/s41586-024-08445-2. Source for the single-shot parity readout, the flux $h/2e$-periodic bimodal quantum-capacitance signal, the roughly one percent assignment error, and the quoted sentence that the measurement "does not unequivocally distinguish between MZMs in the topological phase and fine-tuned low-energy Andreev bound states in the trivial phase."
  3. D. I. Pikulin, B. van Heck, T. Karzig, E. A. Martinez, B. Nijholt, T. Laeven, G. W. Winkler, J. D. Watson, S. Heedt, M. Temurhan, V. Svidenko, R. M. Lutchyn, M. Thomas, G. de Lange, L. Casparis, and C. Nayak, "Protocol to identify a topological superconducting phase in a three-terminal device," arXiv:2103.12217 (2021). The original statement of the topological gap protocol, designed to screen out false positives such as trivial Andreev bound states.
  4. H. F. Legg, "Comment on 'InAs-Al hybrid devices passing the topological gap protocol'," arXiv:2502.19560 (2025). The earlier critique, aimed at Microsoft's 2023 study that first reported devices passing the TGP: the protocol's definitions of "gap" and "topological" differ between the publication and the released code, and its outcomes are sensitive to analysis choices such as the magnetic-field range, bias range, resolution, and cutter-gate settings, rather than being intrinsic to the device.
  5. Microsoft Quantum, "Reply to: On the robustness of topological gap detection via transport" (Matters Arising), Nature 654, E27-E28 (2026). DOI: 10.1038/s41586-026-10568-7. Source for the reply: the argument that the TGP was only a tuning tool while the radio-frequency capacitance signal carries the claim; that a gapless system could not sustain the stable, bimodal $h/2e$-periodic signal because crowding near zero energy would collapse the oscillation; that the size of the capacitance shift indicates a well-developed gap; and that the critique offers no alternative model reproducing the full signal.
  6. The Quantum Investor's Notebook, companion company story, Majorana 2: The Quantum Bet Microsoft Can Afford to Lose.
  7. Microsoft Quantum, "20 Second Parity Lifetime in an InAs-Pb Tetron Device," arXiv:2606.03884 (2026). Company preprint, cited as the company's own claim rather than independently verified: replacing aluminum with higher-gap lead and a reported parity lifetime of about twenty seconds.
  8. Microsoft Quantum, "Distinct Lifetimes for X and Z Loop Measurements in a Majorana Tetron Device," arXiv:2507.08795 (2025). Company preprint, cited as the company's own claim: single-shot measurements of both parity (X and Z) loops of a tetron, with reported assignment errors of about 16% for the X loop and about 0.5% for the Z loop.