Microsoft Doubles Down On Topological Qubits With Majorana 2 Chip

Majorana 2’s Improvements Over Majorana 1

Microsoft published its first quantum roadmap at the same time it announced Majorana 1 in February 2025. The experimental processor was built on an aluminum and indium arsenide material stack, and Microsoft created that chip with the primary goal of proving that hardware-protected topological qubits were possible. (I’ll go into more detail on topological architecture below.)

Majorana 2 includes major changes in both architecture and components. After determining that aluminum was a major cause of quantum decoherence in Majorana 1, Microsoft replaced the aluminum superconductor with one made of lead in Majorana 2. Lead also provides Majorana 2 with better protection from external interference. Besides keeping quantum data from being corrupted by heat or noise, lead provides superior resilience against strong magnetic fields and helps maintain the topological phase, according to Microsoft.

Microsoft also replaced Majorana 1’s semiconductor with a compound of indium arsenide and indium arsenide antimonide. According to the company, the combination of lead and the InAs/InAsSb compound is responsible for the 1,000x improvement in the ratio between how long the qubit stays coherent and how fast it can be operated or measured. Collectively, the improvements increased coherence time from a few milliseconds in Majorana 1 to about 20 seconds in Majorana 2. By quantum computing standards, that is a massive coherence gain.

Engineering Majorana 2 Using Microsoft Discovery

It was significant that Microsoft Discovery’s general availability and Majorana 2 were jointly announced at Build 2026, because Discovery’s AI-powered research acceleration played a pivotal role in the design of the quantum chip. Discovery enabled teams of autonomous AI agents that helped Microsoft scientists develop complex R&D workflows. It synthesized experimental data that had been siloed for decades. And it automated complex qubit measurements and identified fabrication flaws that would otherwise be unnoticed by humans.

Integrating lead into the hybrid semiconductor stack of Majorana 2 required balancing and implementing hundreds of precise fabrication design parameters. Before researchers physically fabricated stacks, they used Discovery to run high-throughput simulations to determine optimal material compositions, appropriate layer thicknesses and the steps needed for manufacturing. It also helped optimize growth conditions for the new InAs/InAsSb. According to the company, that saved weeks in each iteration of growth, cooling and parameter measurements.

Microsoft Discovery’s autonomous agents were also used to make parallel voltage adjustments and measurements across multiple devices simultaneously. Further time was saved by using Discovery Agents to run atom-by-atom crystalline simulations that predicted impurity during molecular beam epitaxy growth. Microsoft Discovery also iterated materials designs before the actual physical fabrication was performed.

The combination of human expertise and Microsoft agentic AI significantly reduced Majorana 2’s development period, but the company says it has also shortened the timeline for developing future commercial quantum computers. Microsoft now predicts that a fault-tolerant machine will be available sometime in 2029 rather than its original projected date in 2033.

Why ‘Majorana’? The Science Behind Microsoft’s Quantum Chips

In 1937, Ettore Majorana, a brilliant Italian theoretical physicist, proposed a special class of fermion that was its own antiparticle. Fermions are a category of the subatomic particles that are fundamental building blocks of matter. Because a Majorana fermion is its own antiparticle, if two of them make contact, they would instantly annihilate each other. Brilliant as he was, Prof. Majorana could not have known how his work would be adapted for quantum computing in the following century.

There is a solid-state version of these fermions called the Majorana zero mode, and Microsoft has based its long-term quantum strategy on MZM. While the potential of this hardware-protected topological architecture is immense, the underlying physics demand unprecedented precision at the atomic level. That characteristic makes topological quantum computing one of the most demanding and interesting areas within physics.

The details of topological quantum computing — and the underlying mathematics of topology — lie beyond the scope of this article. The point is that these ideas are applied to physical systems in a way that allows quantum information to be distributed and made resistant to disturbances that would destroy ordinary qubits. Most qubits are highly susceptible to environmental noise, which is why quantum error correction has attracted so much attention from quantum companies. There are various ways to address QEC. In Microsoft’s case, rather than correcting errors after they occur, the company is developing topological qubits that encode information in non-local quantum states that are resistant to noise in the first place. To put it another way, Microsoft’s long-term goal is to create a hardware-efficient path to fault-tolerant quantum computing that takes advantage of the physics of topological qubits.

Key to Microsoft’s topological architecture is the concept of non-locality. This is in contrast to conventional modalities such as superconducting circuits (favored by IBM and Google), trapped ions (IonQ, Quantinuum) and neutral atoms (QuEra and others), all of which encode quantum information in localized physical states such as local circuit currents or individual atomic energy levels. That type of qubit is susceptible to local environmental noise and stray electromagnetism, which can result in high error rates.

By contrast, topological qubits split the quantum information and store it across separate Majorana zero modes to prevent local disturbances from corrupting the protected quantum state. This relies in part on H-shaped nanowire junctions that act like an electronic traffic network. Electrostatic gates are used to dynamically route signals through different paths. By altering the local electrical environment around the topological nanowire, researchers can shift the locations of Majorana quasiparticles. That action, commonly called braiding, allows quantum information to be manipulated through the use of precision voltage control rather than complex microwave pulse timing.

As touched on above, creating and controlling the exotic states in Majorana 2 is extraordinarily difficult because the materials must be engineered at the atomic level. It also requires temperatures near absolute zero, along with precisely tuned magnetic fields and measurement techniques sensitive enough to detect the behavior of individual electrons.

There is still plenty of work to be done, but if these topological qubits can be perfected, they hold the promise of being much more — even orders of magnitude more — reliable than superconducting qubits, trapped-ion qubits or qubits based on other technologies.

Scientific Controversies Around Microsoft’s Quantum Work

Microsoft has spent nearly two decades pursuing its topological quantum strategy. During that time, several controversies have emerged about the related technology. Before we dig into these, it’s worth noting that the scientific knowledge required for this type of quantum computing is so specialized that there are very few researchers in the world who can evaluate claims associated with this project.

To take a handy example, the Majorana 2 program is led today by Dr. Chetan Nayak, a Microsoft Technical Fellow and Distinguished Engineer who serves as Microsoft’s corporate vice president of quantum hardware. Nayak earned his Ph.D. in physics from Princeton under Nobel laureate Frank Wilczek, where the two jointly made a crucial discovery that underpins Majorana zero modes. This is not garden-variety physics.

The turmoil began with a 2018 paper published in the journal Nature that was written by a Microsoft-funded team at TU Delft in the Netherlands led by Dr. Leo Kouwenhoven. That paper provided evidence that the team had detected Majorana conductance in its experiments. In 2021, outside investigators contended that experimental data in the paper had been selectively presented. That complaint ultimately resulted in the TU Delft team’s retraction of the paper. According to Dr. Nayak, Microsoft was not directly involved in the 2018 TU Delft paper or the data-handling problems that led to the paper’s retraction. As he put it in correspondence with me, “Other than [to] provide part of the funding, we were bystanders.”

A TU Delft integrity committee performed its own investigation and concluded that the researchers had not intentionally committed fraud. MIT physicist Dr. Patrick Lee agreed with Delft committee, but he characterized Kouwenhoven’s work on the project as “sloppy.”

The scientific community has recently resurfaced another debate about Majorana technology. The controversy appeared via the recent publication of a Nature Matters Arising commentary by Dr. Henry Legg of the University of St. Andrews. The comments were originally submitted to Nature in March 2025 — a month after the release of Majorana 1. Dr. Legg argued that there was no clear evidence that Microsoft successfully created a system that produced topological gaps. He also contended that once Microsoft’s system was properly tuned, what were supposed to be gaps turned out to be nothing more than random background noise.

In response, Dr. Nayak issued the following statement: “We stand by our results and our roadmap. At the end of the day, success is the delivery of a scalable quantum computer. We are confident in our ability to execute against our roadmap and proud of our continued engagement with DARPA, which moved Microsoft into the final phase of its Quantum Benchmarking Initiative after independently evaluating our results — those in the public realm and proprietary — with a team of highly qualified experts. Skepticism and rigor are hallmarks of the scientific process, which we appreciate and have supported from various academics. We have participated in dialogue, and our thorough rebuttal was accepted and published by Nature. Most importantly, we are already computing with these qubits.”

Microsoft believes that its peer-reviewed rebuttal clearly shows that Dr. Legg has not been able to provide an alternate theory that is consistent with all of the data in the Microsoft paper.

Dr. Nayak’s comments make me think that for proprietary reasons (or possibly at the U.S. government’s request), it is possible that Microsoft hasn’t shared its complete performance data about Majorana 2. While academic reviewers are limited to publicly available information, the DARPA programs that Microsoft participates in operate under completely different frameworks, with validation by elite scientists from U.S. national laboratories who have access to proprietary data from participants. Given that Microsoft has successfully advanced to the final phase of the US2QC program, it’s a good bet that it is privately meeting DARPA’s stringent engineering requirements.

More important might be Dr. Nayak’s startling final comment that Microsoft is already computing with the qubits in question. That was the first time I had seen any indication that Microsoft has used Majorana 2 to process quantum information. If indeed Majorana 2 is currently operational, it means that while academic skeptics were debating the status of topological signatures, Microsoft was already moving Majorana 2 past the conceptual phase. Now we just need to see the proof.

Remaining Work To Be Done By Microsoft

There are a few specific areas where we know more work is needed. For example, Microsoft’s most current preprint only presents Z-type measurements, but a fully functioning topological qubit requires both X and Z parity measurements. Without both measurements, critics argue that no qubits were actually demonstrated — despite impressive coherence times. As a counter to that, Microsoft has developed a new measurement technique demonstrating that quantum particles are spread out across the material rather than remaining trapped in one spot. According to Microsoft, those results are a hallmark of true topological qubits. While the debate by skeptics will likely continue, in my view Microsoft’s new evidence is too strong to ignore.

Microsoft has a number of other near-term research goals to complete before it can achieve utility-scale performance.

• Braiding and measurement-only gate operations have yet to be demonstrated. Both are critical indications of success.
• Two-qubit entangling gates, which are essential for quantum computing, have yet to be demonstrated.
• Cryogenic control wiring for a million-qubit chip must be accomplished without exceeding the system’s cooling capacity.

For Microsoft to achieve its ultimate vision, it will also need the continued assistance of partner institutions such as QuTech in the Netherlands and the University of Copenhagen, which is a global leader in synthesizing the ultra-pure materials needed for precision crystals and hybrid nanowires. Domestically, Microsoft is aligned with the National Science Foundation, the Department of Energy and DARPA.

The Road To Fault-Tolerant Quantum Computing By 2029

With Majorana 2, Microsoft appears to have changed the math, overhead and timelines for building a commercially viable, error-corrected quantum computer. The transition from the aluminum-based Majorana 1 to a lead-based tetron device architecture was more than just a minor hardware refresh. It was a complete overhaul that supported Microsoft’s underlying thesis by increasing the coherence time from milliseconds to 20 seconds.

Microsoft’s long-term vision is to create a single monolithic chip with a million or more qubits. It plans to deliver the service through Azure Quantum using topological processors to handle computationally intractable subroutines in hybrid quantum-classical workloads. Because interconnects within a chip increase latency, signal loss and calibration complexity, Microsoft believes its topological approach will be more efficient than modular, multi-chip designs used by competitors.

Despite the criticisms about the technology described above, I am optimistic about the validity of Microsoft’s roadmap. Beyond the specific technical advances shown in Majorana 2, the integration of Microsoft Discovery’s agentic AI into the chip’s development has shown that it can accelerate Microsoft’s R&D velocity and potentially shorten its commercial timeline. It is reasonable to think that Microsoft Discovery will have even greater capabilities in the future that could accelerate research even more for future generations of Majorana.

Traditional quantum architectures still face significant issues in managing a large enough overhead of physical qubits for quantum error correction. Conversely, Microsoft’s hardware-protected topological core is a single-quantum-chip solution that appears to be highly scalable and that doesn’t require error correction. That gives me more confidence in the company’s projected delivery date sometime in 2029. Meanwhile, I am looking forward to seeing benchmarks that show exactly how far Microsoft has come with Majorana 2.