Figure 2: Majorana articles, published 2017–2018, that were retracted in 2021–2022 [7]–[9].

Figure 3: Simulated energy E spectrum of a semiconductor-superconductor structure as a function of chemical potential μ. The red line indicates the topological regime where MZMs occur. The green line indicates the non-topological regime where ABSs occur [10, Fig. 11(a)].

Figure 4: The Majorana 1 quantum chip [11].

Perhaps due to the heightened media attention and scientific skepticism, Majorana-related articles have since been published with more caveats and cautious wording about the significance of results. All the usual challenges of experimental design apply in the search for MZMs, but here I will raise a Majorana-specific difficulty: the Andreev bound states (ABSs). ABSs can look exactly like MZMs in conductance measurements [10]. Distinguishing by experiment an almost-zero-energy Andreev bound state from a zero-energy Majorana bound state is itself an active area of research. A trivial ABS is non-topological, so does not exhibit quantum braiding, and therefore is unable to support large-scale quantum computing.

In February 2025, Microsoft published an announcement that they had developed “the world’s first topoconductor, a breakthrough type of material which can observe and control Majorana particles” [11]. The announcement was accompanied by a same-day Nature journal article and news report. The news report emphasised skepticism from physicists in the field [12]. While the journal article contained experiment schematics [13], accompanying data was not made public. There was no demonstration that their measurements were indeed from the MZM and not something else, like the trivial ABS. Microsoft caveats in [11] that the published measurements “do not, by themselves, determine whether the low-energy states detected by interferometry are topological.”

Nayak promised to share detailed qubit data at an American Physical Society meeting that took place in March 2025. Sergey Frolov and Vincent Mourik, the physicists whose concerns led to the 2021 Nature retraction, were transmitted the slide data and afterwards released a dissenting co-written statement [14]: “It is impossible from a physics point of view that Microsoft has demonstrated a topological qubit.” Low data and equipment quality, erroneous methods, and logical fallacies in Microsoft’s claims are all cited by Frolov and Mourik as contributing factors. Their physics-based criticisms “have been established in the scientific literature going back to 2014. It is impossible for Microsoft to not be aware of them.”

The Implications

The multiple Majorana retractions take place against a backdrop of a very divided field and increasingly frequent retractions across all of academia. The upward trend in retractions has been well documented [15]. Proposed explanations include increases in fraud, plagiarism, and AI; better technologies for detecting scientific misconduct; lower barriers to entry as publishers compete with each other; and the ‘publish-or-perish’ priority on output volume [16]. All of these pressures reduce the quality of scientific articles.

However, prestigious journals like Nature and Science should have ample submissions to pick from and abundant peer reviewers to catch subpar articles. Perhaps the cachet of the Majorana 1 team—flush with talent, repute, and funding—is what incentivises big-name journals to publish. If they don’t publish Microsoft, another journal inevitably will and subsequently reap the attention. High impact journals recognise the prestige of Microsoft’s quantum computing researchers, despite intense debate and division.

The peer review document from Microsoft’s February 2025 article is public, demonstrating the differences of opinion. All reviewers agreed that the article was not a yes/no answer to the Majorana question, but disagreed about whether that left any other merits worth publishing. Two reviewers thought that the measurement method used by the Majorana 1 team was itself novel and worthy of Nature. Another said the result was merely a benchmark that is already standard practice in the literature: “On a technical level, I already acknowledged a certain level of advance, but I estimated it well below the bar for publication in a high-impact journal like Nature.” Eventually, the article was published with support from two out of four reviewers [17].

The Majorana 1 team’s guardedness around their experimental data could be out of their control and more about Microsoft protecting their competitive edge. Nayak said the team is “committed to open publication of our research results in a timely manner while also protecting the company’s IP” [11]. The first quantum computer will most likely be introduced to the world as a saleable product, not a knowledge share for giddy physicists. The smart business manoeuvre is to withhold certain details that would allow other scientists to verify (i.e., reproduce) Microsoft’s results. 

This is an uncomfortable tension for many researchers in industry. Traditionally, the spirit of academia has promoted transparency, honesty, and reproducibility. In the physical sciences, reproducibility is especially important to rule out the interference of undisclosed or uncontrolled variables. In 2023, Majorana 1 researchers had to negotiate during the review process about the amount of information they were disclosing in order to successfully publish in Physical Review B. “The intellectual property of the authors' employer has prevented the release of some parameters of the studied devices that may be needed in order to reproduce them,” the journal put out in a subsequent editorial [18]. “As a reflection of the traditional values of the scholarly community, this is not in accordance with the usual norms of the Physical Review journals.”

The problem lies not just with Microsoft, but in an environment where research is dispersing away from universities and toward startups, companies, and governments. When science is protected as a trade secret or classified information, how should we proceed as researchers, publishers, and consumers? Microsoft could very well release the first large-scale quantum computer without releasing the details needed to verify or reproduce the work. Does the ‘how’ matter if the tech just works? To what extent is that good science? We already witnessed a loss of transparency once a novel technology entered the market. OpenAI, the makers of ChatGPT, began as a non-profit organisation seeking to ensure that AI benefits all of humanity. Its models preceding and including GPT-2 were open source. In 2019, the organisation transitioned to a for-profit model and made their subsequent models proprietary. Openness earns trust, but closedness earns profit.

Solid state physicists, Nature, and the Majorana 1 researchers themselves have all stated in one way or another that there is no evidence they have observed MZMs [13]. Microsoft’s statements in [11]—that the “Nature paper marks peer-reviewed confirmation that Microsoft has…been able to create Majorana particles”—is exaggerated at best, fraudulent at worst. Some attach the misleading statements to a bad PR attempt at communicating science, others to stock market manipulation amid a slump in progress. As with the concerns about artificial intelligence overvaluation, it should be noted that quantum computing gains are also consistently over-estimated. For all but a few select applications, like quantum simulation, cryptography, and large-number factorisations, classical computing remains more reliable and practical. Quantum computing requires such refined materials and controlled conditions (most superconducting QCs operate at temperatures colder than deep space) that solving only particular problems can even justify the expense.

The Majorana 1 case study typifies the friction between tech research teams and Big Tech. Competing interests dilute the quality of published work. When intellectual property rights intrude, they threaten traditional scientific norms of transparency and openness. If the project is also a product, advertising takes precedence over objectivity. By impacting publishing, collaboration, funding, and public trust, the sprint to build the first usable quantum computer will only lengthen if these basic scientific principles are forgotten.