The Hype Meets Real-World Physics
Let's get one thing straight. Microsoft's bold 2025 claim that they were "years, not decades" away from a commercial topological quantum computer has run into the harsh reality of peer review. It turns out that the company's highly publicized Majorana particle breakthrough wasn't just built on fragile physics. It was built on basic coding bugs and selective data reporting.
A critique published in Nature by Dr. Henry Legg, a physics lecturer at the University of St. Andrews, has effectively pulled the rug out from under Redmond's quantum roadmap. Legg's paper shows that when you look at the raw transport data Microsoft left out of its original publication, there is no quantum leap. There is only a messy pile of material disorder and some embarrassing Python errors. Talk about a reality check.
As someone who spent years in biophysics labs looking at how physical systems refuse to behave like tidy equations, I find this debate fascinating. It shows the massive gap between what software executives pitch and what is actually happening in the low-temperature dilution refrigerators. Microsoft wanted us to believe they had tamed the Majorana fermion—a particle that serves as its own antiparticle. They claimed they could use its topological properties to build qubits that are immune to noise. The reality is that they couldn't even get their sorting scripts right.
The Flawed Protocol and Omitted Data
To understand where Microsoft stumbled, you have to look at their Topological Gap Protocol. The company developed this automated protocol to detect whether their superconducting devices had transitioned into a topological phase. That transition is the absolute prerequisite for doing any sort of quantum math with Majorana states. If you don't have a clean gap, you don't have a topological qubit. It is that simple.
Legg went through the code and the underlying transport data that didn't make it into the final paper. The transport spectra measure physical charge. In the published paper, everything looked clean. But Legg found that the raw, unpublished data points to significant disorder inside the actual physical devices.
This is a massive problem. In topological physics, material disorder is a killer. It acts like dirt in an engine. It destroys the fragile phase boundary needed to protect quantum states. By omitting this data, Microsoft represented their hardware as clean and compatible with a topological gap when, in fact, it was too messy to support the physics they claimed.
Basic Python Errors That Hid the Truth
The most damning part of the critique isn't a high-minded theoretical disagreement. It is a set of simple programming mistakes that any undergraduate should have caught.
First, Microsoft's plotting software was hardcoded with a filter: zbp_cluster_numbers=[1]. This forced the script to display only the single largest region where the protocol supposedly succeeded. It effectively hid other regions from the phase maps. When peer reviewers asked Microsoft if there were other successful regions in their explored range, Microsoft's team told them they had investigated the only region that passed the protocol. That statement was wrong. If you change the code to zbp_cluster_numbers=[1,2], a second region appears. Microsoft was quite literally blinding itself and its reviewers to its own data.
Second, the code contained a basic array-indexing error. The script reversed a Python array with x[::-1] based on its index position rather than its actual physical bias voltage values. The software antisymmetrized the bias voltage based on where it was stored in memory, not the physical voltage it represented.
It is hard to exaggerate how bad this is. We're talking about a multi-billion-dollar effort to build the world's most advanced computing machine, and it was derailed because someone didn't check how they index arrays in Python. It shows that even the biggest tech companies aren't immune to basic script-kiddie bugs.
Microsoft's Circular Defense
Microsoft technical fellow Dr. Chetan Nayak fired back in Nature, dismissing these issues as minor bugs that don't change the outcome of their research. Nayak pointed to independent evaluations by DARPA and argued that Legg's critique is too focused on transport measurements. He claimed that Legg ignored the capacitance measurements that sit at the core of Microsoft's studies.
But Legg warns that this defense relies on circular reasoning. Microsoft's stance is essentially: We saw a specific capacitance signal, so the prerequisites for that signal must have been met. That is not how rigorous physics works. If your underlying transport data is pointing to disorder and your code is hiding negative results, you cannot just point to a secondary measurement and declare victory.
This isn't Microsoft's only gamble in the space, of which they seem increasingly aware. As we reported in Microsoft's Quantum Leap, the tech giant is also funding neutral-atom systems and electron-on-helium designs. That diversification makes a lot of sense if their topological qubit program is stuck in a loop of bad code and dirty materials.
The Majorana 2 Illusion
Even Microsoft's newer announcements look shaky. In June 2026, details about their Majorana 2 chip emerged, with Redmond boasting that it is 1,000 times more reliable.
But Legg remains unconvinced, pointing out that there is still no proof Majorana 2 is even a single functioning qubit. The preprint on the device fails to report an X-measurement. That measurement is critical for proving a qubit can hold a quantum superposition. Instead, Microsoft's claim of 1,000-fold reliability refers to the lifetime of the classical bit parity.
A classical bit that keeps its state for a long time makes a great memory drive, but it doesn't make a qubit. Until Microsoft can show a real X-measurement and prove they have control over a superposition, Majorana 2 is just another high-level simulator. They are centuries, not decades, away from making this topological dream work.
Sources
- Boffin claims Microsoft's supposed quantum leap does not compute — The Register, June 24, 2026
- On the robustness of topological gap detection via transport — Nature, June 2026
