Microsoft Majorana 2 quantum chip achieves 1,000× reliability gain using Discovery agentic AI during development

Microsoft on June 2, 2026, announced Majorana 2, its next-generation topological quantum chip, alongside the general availability of Microsoft Discovery, the company's agentic AI platform for scientific R&D. The two announcements are linked: Discovery's AI agents played a direct role in developing the chip, automating measurements and identifying fabrication flaws that human researchers had not found at speed.

What's new

Majorana 2 delivers a 1,000-fold improvement in qubit reliability over Microsoft's prior generation. Mean qubit lifetime reaches 20 seconds, with some qubits sustaining coherence for up to one minute. Individual gate operations execute in one microsecond. The chip uses lead rather than aluminum as the superconducting material — a materials substitution that significantly improved electromagnetic shielding and contributed to the reliability gains.

Microsoft now expects to achieve a scalable quantum computer capable of commercially useful computation by 2029, cutting the previously stated timeline roughly in half. Chetan Nayak, Microsoft Technical Fellow, described the generational leap: "We're 1,000 times better" relative to the previous year's progress.

Microsoft Discovery reached general availability simultaneously. The platform lets enterprise customers deploy teams of specialized AI agents — coordinated by human researchers — to manage workflows, automate laboratory measurements, optimize fabrication processes, and propose design changes based on continuous experiment data. A free local version is available as a standalone app requiring a GitHub Copilot account. Zulfi Alam, Corporate Vice President for Quantum, described the platform's contribution to Majorana 2 directly: "Using agentic AI to automate the measurements was a game changer."

Context

Microsoft has been developing topological qubit technology for years, with the first Majorana chip demonstrated in 2025. Topological qubits are designed to be more stable than standard superconducting qubits by encoding information in particle pairs that are inherently error-resistant — a theoretical advantage the Majorana line is built to demonstrate at scale.

The development of Majorana 2 involved continuous deployment of Discovery agent teams across the fabrication and measurement cycle. The agents automated repetitive measurement tasks, flagged anomalous results for researcher review, and proposed material composition adjustments — compressing an experimental timeline that would conventionally require years of iterative human-paced research.

Microsoft Discovery was in enterprise preview for much of 2025, used internally across protein folding, semiconductor material design, and quantum chip fabrication. With GA, it becomes available to external customers on Azure. Early adopter Syensqo is using it to develop next-generation semiconductor manufacturing fluids.

Why it matters

Majorana 2 is a concrete case of agentic AI accelerating physical science at a scale that matters for compute infrastructure. A 1,000× reliability improvement that pulled a quantum computing timeline forward by years — and that was partly enabled by AI agent automation of lab measurement workflows — is the kind of result that justifies the broader enterprise case for agentic R&D platforms.

For the AI infrastructure story more specifically, Microsoft Discovery reaching general availability matters because it packages a workflow previously available only to internal research teams. The quantum chip development is the most credible public proof point for what that platform can do.

For quantum computing, the 2029 target is meaningful. Scalable, commercially useful quantum computation by that date would represent a significant shift in the competitive landscape for cryptography, drug discovery, and optimization problems across logistics and finance. Whether Microsoft reaches that milestone or not, Majorana 2's reliability numbers put the topology-based approach back into serious technical contention.