Most of the quantum computing hardware you hear about — superconducting transmons, trapped ions, neutral atoms — shares a common weakness: the qubits are fragile, and keeping them coherent long enough to compute anything useful requires layering on huge amounts of active error correction. Topological quantum computing proposes something fundamentally different: build qubits whose stability comes from the shape of physics itself, encoded in exotic quasiparticles, so that errors are suppressed by the laws of topology rather than fought off with brute-force engineering.
It’s one of the more mathematically beautiful ideas in the field, and also one of the most experimentally elusive. Microsoft has staked a large part of its quantum computing strategy on this approach, and the story of that bet — including a high-profile retraction — is worth understanding if you want a realistic picture of where this technology actually stands.
The Core Problem Topological Computing Tries to Solve
Every qubit in a conventional platform is vulnerable to local noise: a stray photon, a fluctuating electric field, a thermal phonon, a cosmic ray strike. Because the quantum information is stored in a local physical degree of freedom (the charge on a superconducting island, the internal state of a single trapped ion), any local disturbance can corrupt it. That’s why current quantum processors need quantum error correction — spreading a single logical qubit’s information redundantly across many physical qubits so that local errors can be detected and corrected without directly measuring (and thus destroying) the encoded quantum state.
Topological quantum computing takes a different approach at the physical level: instead of correcting errors after the fact, it tries to make the qubit’s encoded information inherently insensitive to local noise, by storing that information non-locally, in a global topological property of a many-particle quantum system, rather than in the local state of any single particle.
What Anyons Actually Are
To understand why this works, you need a bit of background on particle statistics. In three-dimensional space, quantum particles come in exactly two flavors: bosons (like photons), which are unaffected by swapping two identical particles, and fermions (like electrons), which pick up a factor of $-1$ when two identical particles are exchanged. This binary classification is a consequence of the topology of three-dimensional space — specifically, the fact that the path traced out by swapping two particles twice can always be continuously shrunk to nothing.
In two dimensions, this topological argument breaks down. Swapping two particles twice in a 2D plane is not generally equivalent to doing nothing, because the exchange paths can wind around each other in ways that can’t be undone by continuous deformation. This opens the door to particles with statistics that are neither purely bosonic nor purely fermionic — particles that can pick up an arbitrary phase, or in more exotic cases, undergo an actual transformation of their combined quantum state, when exchanged. Frank Wilczek coined the term anyon for these particles in 1982, capturing the idea that “any” statistics, not just the two familiar ones, become possible in two dimensions.
There are two broad classes relevant to quantum computing:
Abelian anyons pick up a complex phase factor $e^{i\theta}$ when two of them are exchanged, where $\theta$ can be any value (not just $0$ for bosons or $\pi$ for fermions). This is interesting physics, but on its own, abelian anyons aren’t sufficient to build a universal quantum computer — exchanging them only ever produces phase factors, which isn’t enough computational richness.
Non-abelian anyons are the ones that matter for topological quantum computing. When you exchange two non-abelian anyons, instead of just picking up a phase, the quantum state of the whole system transforms according to a matrix operation — and critically, the order in which you perform a sequence of exchanges matters (the operations don’t commute, hence “non-abelian”). This means that braiding non-abelian anyons around each other in different sequences and patterns can implement genuinely different unitary quantum gates.
Encoding a Qubit Non-Locally
Here’s the part that makes this approach so appealing from a fault-tolerance perspective. In the leading theoretical proposal for topological quantum computing, information isn’t stored in the state of a single anyon — it’s stored in the collective, non-local state of a group of anyons, specifically in a shared quantum number related to how they would fuse together if brought close enough to interact.
Because this encoding is a global, collective property distributed across widely separated anyons, no local perturbation — a stray photon hitting one small region of the material, a local charge fluctuation — can, by itself, change the encoded information. To corrupt the qubit, you’d essentially need to move one of the anyons all the way around another in an uncontrolled way, which is a much less likely event than a random local noise fluctuation. This is the entire appeal of the topological approach in one sentence: error protection is built into the physics of the encoding itself, rather than bolted on afterward through active error-correction circuitry.
Braiding: The Computation
Computation in this scheme happens by physically moving anyons around each other in two-dimensional space, tracing out braid patterns over time. Because the anyons’ positions in space, over time, trace out literal braids (in the topological, knot-theory sense of the word), the sequence of gates you apply is determined entirely by the topology of the paths the anyons follow — not by their precise timing, speed, or exact trajectory shape. Small wobbles or imprecisions in exactly how you move an anyon don’t matter, as long as the overall braid pattern — which anyon went around which, in which direction — is not disturbed. This is fundamentally different from, say, a superconducting qubit gate, where the precise duration and amplitude of a microwave pulse directly determines gate accuracy.
Mathematically, a sequence of braiding operations is represented by a product of matrices (typically drawn from a mathematical structure called a braid group representation), and the resulting unitary transformation applied to the encoded qubit state is:
$$U = B_n B_{n-1} \cdots B_2 B_1$$
where each $B_i$ represents one elementary braiding (exchange) operation. For certain types of non-abelian anyons — most notably a hypothetical or theoretical particle called a Fibonacci anyon — sequences of braids have been mathematically proven to be capable of approximating any unitary gate to arbitrary precision, which means a topological quantum computer built from Fibonacci anyons would, in principle, be universal for quantum computation using braiding alone. Other candidate anyon types, notably Majorana zero modes, are not universal through braiding alone (their braid group representations are too limited) and require supplementing braiding operations with additional non-topological gates to achieve universality — an important nuance that’s often glossed over in popular coverage.
Majorana Zero Modes: The Experimental Frontier
Anyons aren’t naturally occurring fundamental particles like electrons — they’re emergent quasiparticles that arise from the collective behavior of many interacting particles in specially engineered materials, typically at very low temperatures, and often in the presence of strong magnetic fields. This is where topological quantum computing shifts from elegant mathematics to extremely difficult condensed matter physics.
The leading experimental target has been the Majorana zero mode — a quasiparticle predicted to emerge at the ends of certain one-dimensional semiconductor-superconductor hybrid nanowires (typically indium antimonide or indium arsenide nanowires coupled to a superconductor like aluminum, in the presence of a magnetic field). Majorana zero modes are named after Ettore Majorana, who in 1937 proposed the theoretical possibility of a fermion that is its own antiparticle. In the condensed matter context, a Majorana zero mode isn’t itself a fundamental particle, but rather an emergent quasiparticle excitation that behaves, in relevant respects, like Majorana’s proposed particle.
Microsoft has pursued this specific physical platform for over a decade as the centerpiece of its quantum computing strategy, on the theory that if you can realize and control Majorana zero modes, you get intrinsically fault-tolerant qubits — the so-called topological qubit — without needing to layer on the same scale of active quantum error correction that superconducting and trapped-ion platforms require.
The 2018 Retraction and What It Means
This is worth being direct about, because it’s a genuinely important case study in how experimental claims in this space should be scrutinized. In 2018, a team including Microsoft-affiliated researchers published a paper in Nature claiming to have observed strong experimental evidence for Majorana zero modes in a nanowire device. The claim was significant enough that it was treated as a milestone for the entire field. In 2021, following renewed scrutiny of the underlying data, the paper was retracted after other researchers and eventually the original authors themselves concluded that the observed signal could be explained by mundane device effects rather than genuine Majorana physics — essentially, the data had been selectively presented in a way that supported the desired conclusion without adequately ruling out simpler explanations.
This episode is a useful reminder that condensed matter signatures consistent with exotic quasiparticles are notoriously difficult to distinguish from ordinary, boring explanations (disorder-induced states, quasi-particle poisoning, and other artifacts), and that claims in this specific corner of physics deserve an unusually high bar of scrutiny before being taken as established.
Microsoft has continued the research program since, with a more cautious and methodologically rigorous approach to validating Majorana signatures, including a widely discussed 2025 announcement of a chip design (“Majorana 1”) that Microsoft describes as based on a topoconductor material stack intended to host and control Majorana zero modes. As with earlier claims in this area, independent replication and sustained scrutiny from the broader condensed matter physics community are the appropriate standard to judge such announcements by, rather than press releases alone.
Why This Approach, If It Works, Would Be a Big Deal
The theoretical payoff of topological quantum computing is substantial enough to explain why serious institutions keep investing in it despite the experimental difficulty. If non-abelian anyons with sufficiently strong topological protection can be reliably created and controlled:
- Dramatically reduced error-correction overhead. Instead of needing hundreds or thousands of physical qubits to encode a single reliable logical qubit (as current superconducting and trapped-ion error-correction schemes require), topologically protected qubits might need far fewer physical resources per logical qubit, because much of the protection is already built into the physics.
- Gate errors suppressed exponentially with distance. The probability of an error scales with how “close” anyons come to accidentally braiding with each other, which in a well-designed device can be made exponentially small by simply keeping anyons well-separated — a very different error-scaling behavior from other platforms.
- Intrinsic robustness to a specific, important class of noise. Local perturbations that would flip a conventional qubit simply don’t have a mechanism to corrupt topologically encoded information, provided the anyons stay well-separated and undisturbed.
Why It Remains Unproven at Scale
No unambiguous, reproducible demonstration of non-abelian anyon braiding for computation yet exists in a solid-state platform at the level of maturity that transmon or trapped-ion qubits have reached. This is the single most important caveat to attach to any discussion of this technology. As of now, this remains fundamentally a research and materials-science program rather than a deployed computing technology.
Material quality requirements are extremely demanding. Realizing clean Majorana zero modes requires nanowires and superconductor interfaces nearly free of disorder — impurities, interface roughness, and unwanted electronic states can all mimic or mask the signatures researchers are looking for, which is exactly what caused the 2018 retraction.
Operating temperatures are extremely low, generally requiring dilution refrigeration comparable to or more demanding than superconducting qubit platforms, undercutting one of the theoretical appeals of the approach (some early proposals suggested topological protection might reduce cooling requirements; in practice, current implementations still need extreme cold).
Universality gaps. As mentioned above, Majorana-based topological qubits are not universal through braiding alone — additional, non-topologically-protected operations are needed to complete a universal gate set, which reintroduces some of the error sources topological encoding was meant to avoid, at least for those specific operations.
Established Physics vs. Speculative Engineering
It’s worth being precise about what’s solid and what’s not. The theoretical framework — anyons, braid group representations, the mathematical case for topological protection and universality with Fibonacci anyons — is well-established, peer-reviewed theoretical physics, unambiguous and broadly accepted within the field. What remains genuinely open is whether any physical material system can reliably host, detect, and controllably braid non-abelian anyons at a scale useful for computation. Fractional quantum Hall systems have shown experimental evidence consistent with abelian and, in some more contested and heavily scrutinized cases, possibly non-abelian anyonic statistics, but not yet in a form suitable for scalable computing. Majorana-based semiconductor nanowire platforms remain an active, difficult, and historically fraught experimental program.
Fractional Quantum Hall Systems: The Other Experimental Path
Semiconductor nanowires hosting Majorana zero modes aren’t the only physical system researchers have investigated for anyonic behavior. The fractional quantum Hall effect — observed in two-dimensional electron gases confined at the interface of certain semiconductor heterostructures, subjected to strong magnetic fields and cooled to very low temperatures — was historically the first physical setting where genuine experimental evidence for anyonic quasiparticle behavior emerged, well before the Majorana nanowire program existed. At certain fractional filling factors (most famously the $\nu = 5/2$ state), theoretical work has predicted the emergence of non-abelian anyons described by a mathematical framework called the Moore-Read (or Pfaffian) state, which would, if confirmed, be directly relevant to topological quantum computation.
Experimental evidence for anyonic statistics in fractional quantum Hall systems has grown more convincing over the past decade, including interferometry experiments designed to directly probe the statistical phase picked up when quasiparticles are exchanged. But translating this into a usable computing platform faces its own serious obstacles: fractional quantum Hall systems require extremely high-quality semiconductor heterostructures, very strong magnetic fields (often several teslas), and millikelvin temperatures, and the specific $\nu = 5/2$ state believed to host non-abelian anyons is comparatively fragile and difficult to stabilize and probe with the precision needed for reliable braiding operations. As with Majorana nanowires, this remains a genuinely active experimental physics research program rather than a computing platform anyone can currently build a processor from.
How Topological Protection Would Change the Error-Correction Calculus
It’s worth spelling out concretely what topological protection would mean for the resource overhead of fault-tolerant quantum computing, since that’s ultimately the practical payoff being chased here. Current leading approaches to quantum error correction on non-topological platforms, such as the surface code used in superconducting and trapped-ion error-correction research, require encoding a single logical qubit’s worth of protected quantum information across a two-dimensional lattice of many physical qubits — commonly cited estimates run from several hundred to well over a thousand physical qubits per logical qubit, depending on the target logical error rate and the physical error rate of the underlying hardware. This overhead is the single biggest obstacle standing between today’s noisy processors (with qubit counts in the hundreds to low thousands) and a genuinely useful fault-tolerant machine, which would likely need many hundreds of well-protected logical qubits.
A topological qubit, by contrast, would ideally need a much smaller ratio of physical resources to protected logical information, precisely because a meaningful part of the error suppression comes from the physics of the anyons themselves — their separation in space — rather than from an actively monitored, syndrome-measured error-correcting code layered on top of otherwise fragile qubits. This is the concrete engineering promise that keeps the research program funded despite decades of experimental difficulty: if it works, it could substantially shrink the physical qubit count needed to reach a given logical qubit count and target error rate, potentially by an order of magnitude or more relative to surface-code approaches on non-topological hardware. Whether that promise survives contact with real materials science, at the scale needed for a genuinely useful processor, remains the open question the entire field is still working to answer.
The Honest Summary
Topological quantum computing is the platform where the gap between theoretical elegance and experimental reality is widest of any major approach currently being pursued. The math is genuinely beautiful and well-understood: two-dimensional physics permits particle statistics that three-dimensional physics forbids, and those statistics can, in principle, be harnessed to build qubits that are fault-tolerant almost by definition. But “in principle” is doing a lot of work in that sentence. Unlike superconducting or trapped-ion qubits, which you can log into and run circuits on today, no one has yet built a topological qubit that a cybersecurity or engineering professional could evaluate in production. If the materials science problem gets solved, it could meaningfully change the resource requirements for fault-tolerant quantum computing. Until then, it’s the platform to watch rather than the platform to plan around.