Microsoft announced the general availability of Microsoft Discovery, its Azure-based platform for deploying autonomous AI agent teams in scientific and engineering R&D workflows. Alongside the GA, Microsoft unveiled Majorana 2, a next-generation topological quantum chip whose 1,000-fold improvement in reliability over its predecessor was achieved in part through Discovery's agentic AI capabilities. The company now expects to deliver a scalable quantum computer by 2029, cutting its original timeline in half.
Microsoft Discovery lets organizations deploy teams of specialized AI agents that reason over large knowledge bases, generate hypotheses, optimize experiments, validate results, and learn in a continuous loop. Built on Azure infrastructure, the platform includes a Discovery Engine for multi-agent research workflows, integration with Azure HPC for compute-intensive simulations, and enterprise-level security, governance, and compliance controls.
The GA announcement highlights that Discovery Engine outputs include confidence scoring and cited research findings, making agent results reviewable and traceable rather than opaque. The platform was shaped by four production R&D requirements: workflows must remain reproducible, outputs must be reviewable, proprietary knowledge must be governed, and agentic systems must fit into the operating model of R&D organizations. A free desktop app in early preview works with a GitHub Copilot account, lowering the barrier for smaller research teams and individual researchers.
Majorana 2 is the most concrete demonstration of what the platform enables. The quantum team used Discovery's agents to manage fabrication workflows, automate measurements, optimize the materials stack, pinpoint previously unnoticed flaws in qubit manufacturing, and correlate patterns across nearly two decades of experimental data in many different formats. Chetan Nayak, Microsoft Technical Fellow, said:
Agentic AI has permeated almost everything we do. It's just become a very natural part of our workflow. The agents can accelerate things as much or as little as you want. It can be as simple as pulling information together and summarizing it, or it can go further by synthesizing it or generating an interesting hypothesis.
The chip itself represents a significant advance over Majorana 1, which InfoQ covered at its introduction in early 2025. According to the technical paper posted to arXiv, Majorana 2 switches from an aluminum to a lead superconductor, which shields qubits from cosmic disturbances. The result: a mean qubit lifetime of 20 seconds, with instances lasting as long as one minute, compared to the microsecond lifetimes typical of other approaches. Operations execute in one microsecond and each qubit measures 1/100th of a millimeter. Zulfi Alam, Corporate VP for Quantum at Microsoft, explained how AI changed the experimental process:
Finding the exact recipe, the right amount to put to get the desired energy structure, requires a lot of experimentation in the old world order. In the new world order, through simulations, you can see where the highly probable target is. And then with that knowledge, you ideally only have to experiment once.
Early GA customers include Pacific Northwest National Laboratory, working on energy storage and biosystems engineering with self-driving scientific workflows that connect AI agents to laboratory automation. Syensqo is developing next-generation fluids for semiconductor manufacturing.
For practitioners, the platform question is whether agentic AI for R&D follows the same adoption pattern as agentic AI for software engineering: specialized agents coordinated by an orchestrator, operating within governed boundaries, with humans setting direction and reviewing results. Discovery's architecture mirrors that pattern, with Copilot as the orchestrator, specialized agents for different research tasks, and the Discovery Engine managing the iterative loop.
Microsoft Discovery is available now on Azure. The Discovery app preview is a free download.