Microsoft Used AI to Design Its New Quantum Chip. Now It Wants to Sell You That AI.

Microsoft built its new quantum chip with AI agents — and is now selling those agents to other researchers. The tool that designed the hardware is the product.

The company unveiled Majorana 2 at Build 2026 on Tuesday, a quantum processor it claims runs qubits 1,000 times more reliably than its predecessor, with some sustaining their quantum state for as long as a minute. Microsoft now says it can build a commercially useful quantum computer by 2029, roughly half the time it estimated a year ago. The topological gap, a key parameter for qubit stability, more than doubled. Operation speed held at one microsecond. Quantum Computing Report

But the more revealing detail was what Microsoft buried in the announcement: the chip's material stack was designed, in part, by the company's own agentic AI system. Microsoft Discovery, the platform Microsoft is now selling to pharmaceutical companies, materials scientists, and other researchers, managed fabrication workflows, identified flaws in material candidates, and proposed new stack configurations. The tool built the thing. Now Microsoft wants to sell you the tool.

"We need to make improvements each year that will get us closer to delivering a computer that we believe will have massive commercial and societal value," said Chetan Nayak, Microsoft technical fellow. "We have got to keep marching to that roadmap, but where are we relative to last year? We are 1,000 times better."

The new material stack swaps the aluminum used in Majorana 1 for lead and antimony, a combination that Microsoft says better shields fragile quantum states from interference. Nayak described the reliability improvement as roughly analogous to a phone battery that instead of dying in a day could run for three years.

The claims are concrete and the timeline compression is real. But the history here is complicated.

Microsoft has been working toward topological qubits based on Majorana quasiparticles for more than two decades, betting that a form of quantum hardware resistant to local errors could scale past what conventional approaches allow. The company has poured more than $1 billion into the effort. Science That persistence has produced results, but also a track record of contested claims. Microsoft published a technical paper on the Majorana 2 architecture concurrently with the announcement. Microsoft Technical Paper

In August 2025, the journal Science corrected a 2020 paper by Microsoft-sponsored researchers that had reported early Majorana signatures. The correction followed a university investigation that found no misconduct, though critics had alleged data cherry-picking. Science Separately, a team from Australian universities published research arguing that decoherence in Majorana qubits remains shorter than the time required to complete measurement operations, a fundamental limitation Microsoft disputes. HPCwire

The independent physics community has not confirmed that Majorana 2 exhibits true topological qubits, as opposed to artifacts that resemble them. Microsoft points to its inclusion in the final phase of DARPA's Underexploited Systems for Utility-Scale Quantum Computing program as external validation. Microsoft That is not nothing. It is also not peer review.

Nayak said Discovery has become a natural part of how his team operates. "The agents can really accelerate things as much or as little as you want. It can be as little as pulling information together and summarizing it, or it can go further down the road of synthesizing it more or generating an interesting hypothesis."

Microsoft is betting that the same AI-driven materials discovery that produced Majorana 2 can be productized for other high-stakes research. Discovery is now generally available for enterprise customers, with a free desktop preview for individual users linked to GitHub Copilot. The company cited early use cases including Syensqo developing next-generation fluids for semiconductor manufacturing.

The recursive structure here is worth sitting with. Microsoft is proposing that the AI which accelerated its own quantum hardware program is now ready to accelerate yours. The claim is internally consistent. Whether it survives contact with real R&D workflows is an open question.

For now, the quantum hardware claims stand alongside the AI claims, both awaiting independent confirmation. Microsoft has a history of announcing bold results before outside researchers can verify them. The gap between what the company says it built and what the field agrees it built has been, over twenty years, a persistent feature of this story.

What changed Tuesday is not just the hardware. It is the framing. The quantum chip is no longer only a quantum story. It is also a demonstration that Microsoft Discovery can design the physical substrate for its own successor systems, and sell that capability to everyone else. The tool did not just accelerate the research. It became the product.