Microsoft has unveiled Majorana 2, a new quantum chip the company says is far more reliable than its predecessor and marks another step toward building a scalable quantum computer.
The company said the chip, developed with help from Microsoft Discovery, uses a new materials stack that delivers qubits that are 1,000 times more reliable than before. Microsoft said the qubits now have a mean lifetime of 20 seconds, with some lasting up to one minute, compared with the microsecond lifetimes common in other approaches. The company said the advance moves it closer to a system it hopes will be commercially useful by 2029, a timeline it has now pulled forward by several years.
Microsoft framed the progress as part of a broader push to use agentic AI in scientific research. The company said its quantum team has been using Microsoft Discovery, a platform for research and development that combines specialized AI agents with human oversight, to manage workflows, automate measurements, optimize fabrication and spot issues in the chip-building process. Microsoft also said a free local Discovery app is now available in preview for individuals with a GitHub Copilot account.
According to Microsoft, the Discovery platform helps researchers deploy AI agent teams that can reason over large amounts of information, propose hypotheses, validate theories and iterate through experiments. The company said the same tools have been used by its quantum scientists and engineers to handle the complex, multi-stage work involved in Majorana 2.
That work includes tracking interactions among software, hardware design, materials and fabrication. Microsoft said the team has nearly two decades of data in different formats, much of it previously siloed. AI agents, the company said, can search across that material and surface patterns that humans might miss. The company also said AI has helped the team handle interdisciplinary research spread across multiple countries and specialties.
Microsoft emphasized that the AI systems provide guidance rather than final decisions, describing the process as “scientist in the loop.”
Majorana 2 builds on last year’s Majorana 1 chip, which introduced a topological superconductor approach aimed at improving stability. Microsoft said the new version changes the materials stack, including a switch from aluminum to lead in the superconductor used to shield fragile qubits from environmental interference.
The company said the change required years of work to manage tradeoffs, but led to major gains in device quality. Microsoft also said the chip is extremely small, with qubit structures measuring about one-hundredth of a millimeter, and can operate in microseconds.
Microsoft described AI as particularly useful in the experimental phase. In one example, the company said an AI agent was built to help automate measurement tasks that had previously taken weeks when done manually. The system reportedly helps map operating conditions in 3D and adjust many variables at once.
The company also said another AI agent helped identify a temperature sensor calibration issue that had affected fabrication data.
Microsoft said the same AI tools used in the quantum program are transferable to other fields, including life sciences, energy, chemicals, manufacturing and consumer goods. The company positioned Discovery as a general-purpose platform for what it calls Frontier R&D, with enterprise controls for security, governance and compliance.
For Microsoft, Majorana 2 is both a hardware milestone and a demonstration of how agentic AI is beginning to shape scientific work in the lab.