Microsoft Claims New Quantum Chip Is 1,000 Times More Reliable Than Its Predecessor

At its Build 2026 conference, Microsoft announced Majorana 2, a next-generation quantum computing chip that the company says is 1,000 times more reliable than the Majorana 1 chip unveiled just 16 months earlier [1]. The qubits on the new chip maintain their quantum state for an average of 20 seconds, with some lasting up to one minute — compared to 1 to 12 milliseconds on the original [2]. Microsoft has simultaneously cut its timeline for a practical quantum computer in half, now targeting 2029 [3].

The announcement has drawn sharp criticism from condensed matter physicists who say the underlying evidence for topological qubits remains unproven, the new preprint has not been peer-reviewed, and the company's quantum division has a documented history of premature claims [4][5].

What Changed: From Aluminum to Lead

The core technical improvement in Majorana 2 is a materials swap. The original Majorana 1 used aluminum as its superconductor. Majorana 2 replaces this with lead, which has a higher superconducting gap — a property that makes it naturally more resistant to disruptions from cosmic rays and other environmental noise that can collapse a qubit's quantum state [2].

Microsoft also updated the semiconductor active region to a combination of indium arsenide and indium arsenide antimonide. The company credits its "Discovery" agentic AI system with identifying the optimal material combinations, claiming AI-driven experimentation accelerated the design process [1].

The result, according to Microsoft: qubit lifetimes roughly 1,667 times longer than the Majorana 1's best performance. Combined with one-microsecond operations and a physical qubit size of 1/100th of a millimeter, the company argues this places it on a path toward commercially useful quantum computing by decade's end [2][3].

What "1,000 Times More Reliable" Actually Means

Microsoft's reliability claim refers specifically to qubit coherence time — the duration a qubit can hold its quantum state before decohering (losing its information to the environment). On Majorana 1, the best coherence times were around 12 milliseconds. On Majorana 2, the mean is 20 seconds [2].

This is not a measure of gate fidelity (how accurately a quantum operation is performed), logical error rates (how often a computation produces a wrong answer after error correction), or computational throughput. It is a single metric — raw qubit lifetime — and while it is an important one, it does not by itself establish that the chip can perform useful error-corrected computations.

For context, coherence times across platforms vary enormously depending on the qubit technology:

Source: Company announcements & published benchmarks

Data as of Jun 2, 2026CSV

Trapped-ion systems from IonQ and Quantinuum already achieve coherence times measured in seconds to minutes, though they face different bottlenecks in gate speed and scalability [6]. Superconducting qubits from Google and IBM operate much faster but decohere in microseconds to milliseconds [7]. Microsoft's claimed 20-second coherence time for a solid-state chip would be exceptional — if confirmed independently.

The Skeptics: "Nothing in This Preprint Resolves the Fundamental Issues"

The Majorana 2 announcement rests on a preprint manuscript that has not undergone peer review [4]. This is a significant caveat, given that Microsoft's previous Majorana 1 preprint — submitted in early 2025 — also remains unpublished more than a year later [5].

Henry Legg, a physicist at the University of St Andrews, told Science News that "nothing in this preprint resolves the fundamental issues" raised about the Majorana 1 [4]. Legg argues that the switching behavior Microsoft presents as evidence of topological qubits could instead be explained by an electron hopping on and off a quantum dot — a mundane semiconductor artifact, not an exotic topological state [4].

Sergey Frolov, a physicist at the University of Pittsburgh who helped trigger the retraction of Microsoft's 2018 Nature paper, has been consistently critical. He characterized Microsoft's Majorana data as "just noise" and suggested that the eight-qubit Majorana 1 chip "cannot possibly work, given what we saw" [8]. He has described a "sustained pattern of unreliable claims" from Microsoft's quantum division [8].

The core scientific question is whether Microsoft has actually produced and controlled Majorana zero modes — exotic quasiparticles that emerge at the boundaries of topological superconductors and that would, in theory, encode quantum information in a way that is inherently resistant to local disturbances. No experiment has yet met the scientific community's standard of proof for non-Abelian anyons, the class of particles that includes Majorana zero modes [9].

How Topological Qubits Differ — and Why They Matter

Most quantum computers today use one of two approaches: superconducting circuits (Google, IBM) or trapped ions (IonQ, Quantinuum). Both encode quantum information in fragile physical states — electromagnetic oscillations in a superconducting loop, or the energy levels of individual atoms held in electromagnetic traps.

Topological qubits, the approach Microsoft has pursued since 2017, would work differently. Instead of storing information in a single particle's state, a topological qubit encodes information in the collective behavior of many particles — specifically, in the way quasiparticles braid around each other in two-dimensional space. Because this information is stored non-locally, no single local disturbance can corrupt it. The qubit would be, in Microsoft's framing, "reliable by design" [10].

The theoretical advantage is substantial: where superconducting and trapped-ion systems need thousands of physical qubits to error-correct a single logical qubit, a topological approach could need far fewer, because the hardware itself suppresses errors at the physical level [10].

The practical problem: after nearly a decade of work, the existence of the requisite Majorana zero modes in Microsoft's devices has not been conclusively demonstrated to the satisfaction of the broader physics community [5][9].

The 2018 Retraction and Microsoft's Track Record

Microsoft's quantum program has been shadowed by a high-profile failure. In 2018, a team based at a Microsoft laboratory in the Netherlands published a paper in Nature claiming to have observed the quantized conductance signature of Majorana zero modes — a result that, if correct, would have been a major step toward topological qubits [11].

In 2021, the paper was retracted. An investigation found that the researchers had "unnecessarily corrected" data without disclosure and mislabeled a graph in ways that made results appear more conclusive than they were [11]. Frolov, who had obtained the raw data and identified discrepancies, showed that another quantum phenomenon could mimic the observed signatures. The retraction noted that an independent review found evidence of selective data presentation [12].

This history is relevant because the current Majorana 2 claims rest on similar experimental signatures — measurements that the company interprets as evidence of topological behavior, but that critics argue could have prosaic explanations [4][5].

The Competitive Landscape: Where Microsoft Stands

While Microsoft pursues topological qubits, its competitors have been posting measurable results with more established technologies.

Google demonstrated in late 2024 that its 105-qubit Willow processor could perform quantum error correction below the threshold — meaning that adding more physical qubits actually reduced the logical error rate, rather than increasing it. Their distance-7 surface code achieved a logical error rate of 0.143% per correction cycle [7].

IonQ set a world record for two-qubit gate fidelity at 99.99%, the first company to reach the "four nines" benchmark [6].

Quantinuum demonstrated 48 logical qubits from 98 physical qubits in November 2025, and has announced a roadmap to universal fault-tolerant quantum computing by 2030 [13].

Source: Company announcements & published results

Data as of Jun 2, 2026CSV

Microsoft's own Azure Quantum platform currently provides cloud access to hardware from IonQ, Quantinuum, and other partners — an acknowledgment that its own topological hardware is not yet ready for external use [14].

How Many Qubits Would It Take?

The Majorana 1 chip contained 8 physical qubits. Microsoft has not disclosed the qubit count on Majorana 2. Its long-term roadmap targets one million physical qubits on a single chip [15].

The practical thresholds that matter for real-world applications are measured in logical qubits — error-corrected units composed of many physical qubits:

• Breaking RSA-2048 encryption: Estimates range from roughly 1,730 logical qubits using Shor's algorithm (a 2024 study) to fewer than one million noisy qubits in a recent Google Quantum AI estimate [16]. Even optimistic projections require error-corrected logical qubits that no platform can yet produce at scale.

• Drug discovery and materials science: Useful quantum chemistry simulations — modeling molecular interactions for pharmaceutical development — are estimated to require hundreds to thousands of logical qubits with very low error rates, a capability that remains years away on any platform [16].

Microsoft's 8-qubit Majorana 1 and whatever qubit count Majorana 2 carries are far from these thresholds. The company's value proposition rests not on current capability but on the argument that topological qubits will eventually scale with less error-correction overhead than competing approaches.

Research Interest in Topological Qubits

Academic interest in topological qubits grew steadily from 2011 to a peak of 4,577 papers in 2023, according to OpenAlex data [17]. Publication volume has since declined, falling 20.1% from 2024 to 2026's projected total of 3,192 papers.

Source: OpenAlex

Data as of Jan 1, 2026CSV

This decline may reflect growing skepticism about the field's experimental progress, a natural maturation cycle, or both. The peak in 2023 coincided with intensifying debate over Microsoft's claims and broader interest in quantum error correction.

The Post-Quantum Cryptography Clock

Regardless of when fault-tolerant quantum computers arrive, governments are already preparing. NIST finalized its first post-quantum cryptography standards in 2024, and as of January 2026, CISA published a list of product categories where quantum-safe alternatives are commercially available [18].

The UK's National Cyber Security Centre has published migration timelines urging organizations to begin inventorying their cryptographic dependencies now [19]. For U.S. National Security Systems, certain components were required to transition by December 31, 2025, with broader migration deadlines extending to 2030 [18].

Microsoft's Majorana 2 announcement does not change these timelines. The chip is a research prototype, not a cryptographically relevant machine. NIST's guidance has consistently treated the quantum threat on a 15-to-25-year planning horizon, and no single hardware announcement from any company has altered that assessment [18].

Microsoft's Investment and What's at Stake

Microsoft has invested over $1 billion in quantum computing research, according to industry analyses [14]. CEO Satya Nadella has described quantum computing as "the next big accelerator in cloud" and positioned it as central to Azure's long-term strategy [20].

The company's partners include Atom Computing (with whom Microsoft is building what it calls "the world's most powerful quantum machine"), Quantinuum, IonQ, and Rigetti — all accessible through Azure Quantum [14]. Microsoft also participates in DARPA's Underexplored Systems for Utility-Scale Quantum Computing (US2QC) program, which provides government funding tied to demonstrable milestones [3].

The financial and reputational stakes are high. If topological qubits prove to be a dead end — or if the timeline to fault tolerance stretches well beyond 2029 — Microsoft will have spent a decade and substantial resources on an approach that competitors bypassed with more conventional technologies. IBM, Google, and Quantinuum are all targeting fault-tolerant quantum computing within similar timeframes using superconducting or trapped-ion qubits that have already demonstrated error correction below threshold [7][13].

Conversely, if topological qubits deliver on their theoretical promise, Microsoft would hold a fundamental hardware advantage: qubits that require far less error-correction overhead, enabling more efficient scaling to the millions of logical qubits needed for practical applications.

The Bottom Line

Microsoft's Majorana 2 represents a measurable improvement over its predecessor by the specific metric of qubit coherence time. The 1,000x improvement in stability is a materials engineering achievement — switching from aluminum to lead — and the company's use of AI to optimize chip design is a genuinely novel methodology [1][2].

But the deeper question — whether Microsoft has actually created topological qubits — remains unanswered. The preprint has not been peer-reviewed. The scientific community's leading critics say the data is consistent with mundane explanations. And the company's history includes a retracted paper and claims that have not held up to outside scrutiny [4][5][8][11].

Microsoft is asking the world to take a large bet on its interpretation of contested physics. The 2029 timeline for a practical quantum computer depends not just on better materials but on proving that topological qubits are real, controllable, and scalable — proof that, as of June 2026, does not exist in the peer-reviewed literature.