Microsoft’s Majorana 2: Agentic AI shortens quantum development

Microsoft announced it used its agentic AI platform, Microsoft Discovery, to accelerate development of the Majorana 2 quantum chip. The company reported qubits 1,000 times more reliable than Majorana 1, a mean qubit lifetime of 20 seconds and individual qubits that in some tests maintained state for up to a minute. Microsoft updated its internal timeline for a commercially scalable quantum computer from 2033 to 2029. Microsoft Discovery reached general availability this week.

Microsoft identified the single most significant materials change as replacing aluminium with lead for the superconducting component; that choice resulted from conventional materials research rather than an AI recommendation. Microsoft described the role of agentic AI as managing fabrication workflows, automating measurement tasks, integrating nearly two decades of laboratory data and surfacing correlations across diverse datasets.

Engineers said automating qubit measurement reduced time previously spent on manual tuning. An agent built on Microsoft Discovery runs measurements continuously, constructs three-dimensional maps of qubit conditions and adjusts hundreds of voltage parameters in parallel. The measurement process determines whether a semiconductor wire holds an even or odd number of billions of electrons.

Zulfi Alam, corporate vice president for quantum at Microsoft, commented: “As you run AI agents on this data, they’re able to essentially resynthesize and make correlations that we as humans cannot see because no single individual has that much vision across that much data.” He added that automating measurements shortened experimental cycles.

Chetan Nayak, a technical fellow who leads Microsoft’s quantum programme, noted that agentic AI has become integrated across the team’s workflow and framed the 1,000-fold figure as a year-over-year comparison to Majorana 1: “Where are we relative to last year? We’re 1,000 times better.”

Microsoft Discovery combines specialised AI agents with a Discovery Engine for research workflows and includes enterprise security and governance features. The platform is available to enterprise customers and a free Discovery app is in early preview to run locally with a GitHub Copilot account. Microsoft cited early adopters in life sciences, chemicals and materials, energy and manufacturing, and named Syensqo as an example working on fluids for semiconductor production.

Microsoft stated that the 1,000-fold reliability improvement refers specifically to Majorana 1’s qubits and does not serve as a direct benchmark against other quantum architectures.