With a paper published in the 19 February 2025 issue of Nature [
1], Microsoft (Redmond, WA, USA) fanned the flames of its unique vision for quantum computing: a stable, error-resistant qubit based on the Majorana fermion, one of the strangest and most elu-sive particles in physics. The Microsoft Azure Quantum research team’s descriptions of a means to detect the as-yet theoretical par-ticles [
1]—called "an entirely new state of matter" by Microsoft’s chief executive officer [
2]—and a design for a chip powered by them (
Fig. 1) [
3] have refocused attention on the company’s ambi-tion to build a topological quantum computer. The approach—if it works—could potentially leapfrog every other in the field.
"If you can eliminate almost all decoherence, you have nailed it," said Daniel Lidar, professor of engineering and director of the University of Southern California (Los Angeles, CA, USA) Center for Quantum Information Science and Technology. "This is the holy grail of quantum computing."
But whether the new results merit Microsoft’s enthusiasm remains uncertain, as the scientific community appears largely unconvinced of the company’s claims [
4-
8]. As Lidar put it, "So far, there is no consensus that their results have pinpointed the observation of a Majorana."
Controversy regarding Microsoft’s Majorana project first arose with a 2018 Nature paper authored by some of its Azure Quantum researchers, a report retracted in 2021 following scrutiny of the team’s data analysis [
9-
11]. "There is a very high bar for proving the existence of Majorana particles," Lidar said. While the 2018 results were consistent with the observation of a Majorana fer-mion, they were also consistent with other explanations, he said. "In the end, it turned out the other explanations were correct— something similar could be happening with their new results."
In addition, work published in a 2020 "First release paper" in Science, sponsored by Microsoft and conducted primarily by University of Copenhagen (KU) researchers, purported to have observed Majorana particles within tiny "nanowires" [
12]. In 2021, however, in response to concerns regarding "data cherry-picking" raised by researchers at the Jülich Research Center in Jülich, Germany [
13,
14], the publisher placed an "Editorial Expres-sion of Concern" on the paper [
15]. Four years later in August 2025, following the completion of a KU investigation into the matter, Science removed the editorial expression of concern and added an erratum that included data left out of the original paper [
16]. The KU investigation concluded that, although the researchers had used subjective judgment in data selection, their behavior did not constitute negligence or scientific misconduct and the excluded data did not undermine the paper’s main conclusions; regardless, at least some of the skeptics remain unconvinced [
13,
14].
In the arcane physics of subatomic particles, a dozen elemen-tary fermions have been discovered so far, including six types of quarks and six types of leptons, and several others have been pre-dicted but not yet observed experimentally [
17]. But setting them apart from most other subatomic particles, which have distinct antiparticles (e.g., electrons and their positrons), Majorana fer-mions are unique particles theorized to exist as their own antipar-ticles [
18]. Given this property, the notion of using Majorana particles has intrigued quantum computing experts for more than two decades [
1]. Majorana particles, first proposed by Italian physi-cist Ettore Majorana in 1937, are theorized to emerge in topologi-cal superconductors, whose overall structure or shape provides protection from environmental noise, unlike conventional super-conductors [
18].
Microsoft’s Majorana approach relies on an intricate hardware setup composed of semiconducting indium antimonide nanowires wrapped in a superconducting aluminum shell. The nanowires are precisely shaped and aligned to reduce environmental noise that can destroy delicate quantum states. At sub-Kelvin temperatures, the aluminum enters a superconducting phase. Under the right voltage and magnetic field, the researchers observed what they say are signatures of Majorana zero modes—localized states that behave like Majorana fermions—emerging at the ends of the wires [
1,
18].
"Majorana zero modes are very difficult to measure," said Itay Hen, an associate professor (research) in physics and astronomy at the University of Southern California who develops quantum simulation algorithms. "The fact that they are robust states and not easily destroyed also means that they are hidden away from the environment."
If the signatures measured as reported truly represent Majorana zero modes, the hardware that Microsoft has built could form the basis of highly robust qubits [
2], immune to the noise that plagues most other quantum computing architectures. These architectures are typically comprised of hundreds of physical qubits that pro-duce a single logical qubit, in some cases with at least 99% of the physical qubits taking on the role of error correction [
19]. "If a Majorana qubit lives up to the hype, fundamentally we would only need a hundred qubits for a useful machine," said Bert de Jong, senior scientist and head of computational sciences at Lawrence Berkeley Laboratory in Berkeley, CA, USA.
Majorana particles’ posited advantage lies in their topology, the spatial relationships among their constituent parts. Information stored in a topological qubit is non-local, almost like tying a knot in a rope. If the knot stays tied, it does not matter how the rope is bent or stretched—the information is preserved. "The beauty of topological quantum computing, particularly using fermions, is that it is extremely stable against environmental disturbances," Lidar said. "The qubits are inherently robust—if anybody can con-vincingly demonstrate this, it will be a major step forward."
The Microsoft team claims to have demonstrated a key feature of Majorana systems, non-local tunneling, in which quantum informa-tion is stored across spatially separated particles consistent with a topological phase [
1]. But critics say the evidence remains ambiguous. "Their paper does not actually claim that they observed Majorana fer-mions, let alone manipulated them," Lidar said. "It was more of a blue-print and a test of an experimental spectroscopy device that could establish a signature of the fermions’ presence."
The growing scrutiny includes published critiques of various aspects of the Microsoft team’s research. For example, a researcher from the University of St Andrews (Scotland, UK) authored a pre-print in March 2025 raising concerns about the test that Microsoft uses to look for Majoranas [
20]. Known as the topological gap pro-tocol (TGP), the test is not mentioned in Microsoft’s February 2025 Nature paper. But the company has subsequently indicated that it created the topological qubits using the TGP [
6]. A team at the University of New South Wales (Sydney, Australia) published another preprint in June 2025 suggesting that the Majorana parti-cles’ decoherence time is too short to support their use as qubits and that significant materials breakthroughs would be needed to deliver adequate decoherence times [
21], contentions that Micro-soft has disputed [
22].
In July 2025, Microsoft published another preprint describing a device with the ability to fully control and read out a topological qubit’s state [
23]. The paper also reports decoherence times suffi-ciently long to allow quantum computations. But again, other experts note that these measurements could be consistent with a non-topological system, and there is no definitive evidence pre-sented for having observed a Majorana fermion [
24].
The good news, Lidar said, is that once Microsoft researchers have definitively observed Majorana fermions, the company will essentially have full logical qubits in hand. But building a topologi-cal qubit is only the beginning. The real test will be entangling multiple Majorana qubits without losing the noise resistance that gives them their appeal. That may be the biggest challenge for the Microsoft team, Lidar said. "It is not clear that they can connect multiple qubits together and still preserve the system’s topology because those connections themselves may be prone to environ-mental noise."
Despite the setbacks and skepticism, Lidar sees no evidence that Microsoft is deterred from pursuing a topological quantum com-puter. "If anything, they are doubling down, and that is great," he said. "There is a consensus in the community that it is super impor-tant that this approach continues to be pursued." Hen said he also sees Microsoft’s firm commitment to Majorana particles. "What they are doing is new on every front: The qubits are new, the fab-rication is new, the protocols are new, everything is new and unique to them—it would be very difficult to do other things."
At the same time, though, Microsoft appears to be hedging its big Majorana bet with a hardware-agnostic software stack for quantum computing. "While Microsoft is investing heavily in hard-ware technology, the company is also developing a software stack that will work across platforms," de Jong said. In November 2024, Atom Computing (Boulder, CO, USA) used Microsoft’s Azure Quantum Compute Platform to entangle 24 logical qubits made from neutral atoms, the highest number of entangled qubits observed at the time [
25], now surpassed by Quantinuum (Broom-field, CO, USA) which reported 50 entangled qubits in May 2025 [
26]. The Microsoft and Atom teams also created 28 logical qubits from 112 physical qubits with the ability to detect and correct errors while performing reliable computation [
25]. The two com-panies are building a commercial quantum computer for the Novo Nordisk Foundation (Hellerup, Denmark), scheduled to go online by 2027 [
27. Microsoft has also partnered with Quantinuum in the creation of a trapped-ion quantum computer with 12 logical qubits built from 56 physical qubits [
28].
As a bellwether of Microsoft’s hardware progress, Hen and de Jong both pointed to the company’s successful advancement through the US Defense Advanced Research Projects Agency (DARPA) Quantum Benchmarking Initiative (QBI) [
29]. In early February 2025, just two weeks before publication of the Nature paper, DARPA chose the company, along with PsiQuantum (Palo Alto, CA, USA), which uses a photonic approach to encode qubits in single photons traveling through optical circuits, to advance to the final stage of the QBI’s Underexplored Systems for Utility-Scale Quantum Computing program [
30]. "It definitely means something if the government says that you are advancing from stage A to B to C," Hen said.
Microsoft’s Majorana-based hardware approach notably differs from the quantum computing architectures pursued by other industry leaders like IBM (Armonk, NY, USA), Google (Mountain View, CA, USA), and Amazon (Seattle, WA, USA). These companies have been advancing superconducting qubit platforms and focus-ing on improving gate fidelities, coherence times, and error correc-tion through novel engineering [
31]. For instance, IBM’s latest 156-qubit Heron R2 chip uses a hexagonal qubit layout with tunable couplers and other innovations to suppress noise, allowing circuits with thousands of operations to run reliably [
32]. The company hopes to build a machine with 200 logical qubits by 2028, with a cloud-accessible version by 2029 [
33]. Google’s recently described Willow chip demonstrated that adding more qubits in a surface code can reduce error rates exponentially, hitting a key threshold for logical qubit improvement [
34]; the company is eyeing a machine with 100 or more logical qubits by 2028 [
35]. Amazon’s Ocelot uses a more efficient error correction scheme that requires nine physical qubits per logical qubit [
36].
Other platforms, such as trapped-ion systems, may offer several advantages over superconducting-based quantum computers, including higher accuracy quantum operations, longer coherence times, all-to-all connectivity, and easier scaling [
37]. Leaders in this approach include Quantinuum and IonQ (College Park, MD, USA) [
38,
39]. IonQ says it will have built a trapped-ion system with 2 million physical qubits yielding 40 000 to 80 000 error-corrected logical qubits by 2030 [
31].
However, regardless of the immense investment supporting these myriad platforms and various milestones surpassed [
34,
40,
41], the hoped for and much hyped great promise of quan-tum computing still appears to remain as elusive as the stealthy Mayorana. What people are really interested in is a breakthrough in terms of a practical application, Hen said. "Honestly, I am some-what pessimistic about whether we will ever see something," he said. "But if we do, it will first be in the field of quantum simula-tions to explore quantum materials, which could help with drug design or materials science."
PDF
(386KB)
