How topological qubits work
How Majorana zero modes could protect a qubit by design, how Microsoft builds and measures its devices, why the evidence is disputed, and what would settle it.
A topological qubit would store quantum information where local noise cannot reach it: spread across pairs of exotic states, called Majorana zero modes, that sit at the two ends of a superconducting nanowire. If it works, much of the protection that other qubits get from error correction would be built into the hardware. Microsoft leads this approach with its Majorana 1 and Majorana 2 devices. It is the least mature and most contested of the six modalities: as of October 2026 no qubit operation has been published, and whether the devices reach the required topological phase is disputed.
This article is part of a series on the main ways to build a qubit. The overview compares all six side by side.
How the qubit would form
In 2001 Alexei Kitaev showed that a chain with a special kind of superconductivity can leave one unpaired "half-electron", a Majorana mode, at each end. Together the two end modes form one ordinary electron state that can be empty or filled at almost no energy cost. Because the two halves sit micrometers apart, a disturbance that touches only one end can neither read nor flip that state. This non-locality is the hoped-for topological protection.
No natural material behaves this way conveniently, so proposals from 2010 engineer one: a semiconductor nanowire with strong spin-orbit coupling, such as indium arsenide, made superconducting by a metal layer, in a magnetic field of about 1 to 2 tesla along the wire. Above a critical field, the wire is predicted to become a topological superconductor with Majorana modes at its ends.
A qubit needs four Majorana modes. States of different total parity cannot be superposed, so a single pair cannot hold a qubit; with four modes and the total parity fixed, two states remain, and they form the qubit. Microsoft's tetron joins two topological wire segments with a superconducting backbone. Measuring the parity of the two modes on one wire gives the qubit's Z value, and measuring a pair that spans both wires gives its X value.
The protection is only as strong as the topological gap, the energy that separates the zero modes from other states. Reported gaps are small: about 20 to 30 micro-electronvolts (µeV) in Microsoft's aluminum devices, as measured with its disputed gap protocol, and about 70 µeV in the best fifth of its lead devices, in a 2026 preprint.
How the devices are made
Microsoft grows indium arsenide quantum wells by molecular beam epitaxy and deposits the superconductor in the same vacuum, so the interface is atomically clean. The wires are then defined by etching and gate electrodes in this flat stack, a process compatible with lithography. Its 2026 devices replace aluminum with lead and use a redesigned quantum well on a gallium antimonide substrate, with higher electron mobility and less disorder.
The chips run in a dilution refrigerator at about 10 to 50 millikelvin, with a magnetic field of about 1 to 2 tesla aligned with the wires. Microsoft describes Majorana 1 as eight qubits on a chip designed to scale to a million. Both are company claims, and critics dispute that the devices are qubits yet.
How it would be controlled and read out
Instead of moving Majorana modes around each other, known as braiding, Microsoft's architecture is measurement-only: sequences of parity measurements on pairs of modes carry out the operations without moving anything.
A parity measurement couples a quantum dot to two Majorana modes through tunnel junctions, forming an interferometer loop. The dot's quantum capacitance depends on the pair's parity, and radio-frequency reflectometry reads it in microseconds. In Microsoft's 2025 Nature paper, a single shot reached a signal-to-noise ratio of one in 3.6 microseconds (µs), with about 1% assignment error.
Lifetimes have improved. In a 2025 preprint, the Z parity of an aluminum-based tetron lasted 12.4 milliseconds (ms), while its X parity switched about 1,000 times faster, in 14.5 µs, with 16% assignment error. In the June 2026 Majorana 2 preprint, the Z parity of a lead-based tetron lasted about 20 seconds; X was not reported.
Measurements alone give only part of a universal set of operations, the Clifford gates. The remaining gates are not protected for Majorana modes and need magic-state preparation and distillation, as in other modalities.
Why the evidence is disputed
No experiment has yet shown a signature unique to Majorana zero modes, such as their braiding statistics, or a coherent qubit with stable X and Z measurements. When Nature published Microsoft's 2025 parity-measurement paper, its editors added a note that the results are not evidence of Majorana zero modes in the devices.
Henry Legg of the University of St Andrews has argued that the topological gap protocol, the test Microsoft uses to identify the topological phase, gives different answers under small changes to its settings. His critique, which also identified a code error, was published in Nature as a Matters Arising on June 24, 2026. Microsoft acknowledged the bug, called it minor and stands by its results.
The field's history adds to the caution. A 2018 Nature paper on quantized Majorana conductance, from a Microsoft-funded group at Delft, was retracted in 2021. Similar signals can come from ordinary effects, such as Andreev bound states, disorder or quantum dots.
Supporters point to real engineering progress: high-quality heterostructures, fast single-shot parity readout and parity lifetimes of about 20 seconds (2026 preprint) are measurable achievements, whatever the answer to the topology question.
What would settle it
Microsoft's published roadmap has four stages: a single-qubit device for benchmarking by measurement, a two-qubit device with braiding by measurement, an eight-qubit device that shows an error-correction advantage, and an array for fault tolerance. Passing the first stage means stable X and Z measurements and coherent control of one qubit. Microsoft says its timeline to a scalable machine has moved forward to 2029, a company claim, and with PsiQuantum it is in the final phase of DARPA's evaluation of utility-scale quantum computing.
Other routes to topological protection are being explored. Nokia Bell Labs works with fractional quantum Hall states in ultra-clean gallium arsenide, and QuTech builds Kitaev chains from coupled quantum dots. Quantinuum and Google have created and braided non-Abelian anyons on conventional qubits; those are simulated topological states rather than protected hardware.
Strengths and limits
Potential strengths. Protection built into the hardware, which could reduce error-correction overhead; a unit cell micrometers across; digital control by gate voltages and measurements; lithographic fabrication; and Z-parity lifetimes of about 20 seconds in a 2026 preprint, read out in microseconds.
Limits today. No demonstrated qubit, gate or two-qubit operation; the presence of Majorana zero modes in these devices not established to the field's satisfaction; gates outside the Clifford set still need distillation; small topological gaps, sensitivity to disorder, fields of about 2 tesla and millikelvin temperatures; and a history of retractions and disputed analyses.
Key numbers
Topological gap: about 20 to 30 µeV (aluminum devices, by Microsoft's disputed protocol); about 70 µeV in the best lead devices (2026 preprint).
Parity readout: signal-to-noise ratio of one in 3.6 µs, about 1% assignment error (Nature, 2025).
Z-parity lifetime: 12.4 ms (aluminum tetron, 2025 preprint); about 20 s (lead tetron, 2026 preprint).
X-parity lifetime: 14.5 µs with 16% assignment error (aluminum tetron); not reported for lead.
Qubit coherence, gate fidelity, two-qubit operation: not demonstrated as of October 2026.
Operating conditions: about 10 to 50 millikelvin, with a magnetic field of about 1 to 2 tesla.
Quick answers
What is a Majorana zero mode? A quasiparticle state predicted at the ends of a topological superconductor that behaves like half of an electron. Two of them together store one bit of parity information non-locally.
Has anyone built a topological qubit? Not in a form the field accepts. Microsoft has shown fast parity measurements and long Z-parity lifetimes, but no coherent qubit operation, and whether its devices host Majorana zero modes is disputed.
Why would topological qubits need less error correction? Their information would be stored non-locally, so most local noise could not disturb it, and errors would be suppressed exponentially by the wire's length and topological gap.
Does Anyon Computing build topological qubits? No, our processors use superconducting qubits. Anyons, quasiparticles of two-dimensional systems with exchange statistics unlike those of bosons or fermions, are still central to error correction on superconducting qubits. The surface code, which Google used to show error correction below threshold on a superconducting processor, descends from Kitaev's toric code: repeated stabilizer measurements hold a patch of qubits in a topologically ordered state, a chain of errors leaves anyons at its ends, and the decoder finds those anyons and pairs them up. A logical error needs a chain that stretches across the whole patch. Superconducting processors have also prepared toric-code states and probed their anyons (Google, 2021) and braided non-Abelian anyons, so far without topological protection (Google, 2023; Zhejiang University, Tsinghua University and collaborators, 2024). Decoding has to keep pace with the processor over a low-latency link to classical computers, one reason we couple our processors to GPUs over NVIDIA NVQLink. We took our name from anyons.
Sources
Aghaee et al., 20-second parity lifetime in an InAs-Pb tetron device (arXiv 2606.03884)
University of St Andrews, critique published by Nature (June 2026)
Nature News, Microsoft upgrades controversial quantum chip (June 2026)
Zhang et al., quantized Majorana conductance, Nature 556, 74 (2018), retracted in 2021
Kitaev, fault-tolerant quantum computation by anyons, Annals of Physics 303, 2 (2003)
Dennis et al., topological quantum memory, Journal of Mathematical Physics 43, 4452 (2002)
NVIDIA, NVIDIA introduces NVQLink, press release (October 28, 2025)
About Anyon Computing
Anyon Computing Inc. (also known as Anyon Technologies) is a quantum computing company that provides superconducting modular quantum supercomputers, which it tightly couples with NVIDIA GPUs via NVQLink for data center deployment at scale. Anyon pioneered deploying NVIDIA NVQLink inside a commercial production data center and is among the first superconducting QPU backends integrated with NVIDIA CUDA-Q. Founded by pioneers from Caltech and UC Berkeley. https://www.anyoncomputing.com/
NVIDIA, CUDA-Q and NVQLink are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. Other company and product names may be trademarks of the respective companies with which they are associated.
Media contactAnyon Computing · info@anyoncomputing.com