By Microsoft used AI to design a quantum chip. That chip — Majorana 2, unveiled at Build 2026 — delivers qubits 1,000 times more reliable than its predecessor. Microsoft now says it will have a practical quantum computer by 2029, three years ahead of its previous target. Some of the world’s leading physicists think the entire approach is fundamentally flawed. Welcome to the frontier.
What Actually Changed: Lead, Not Hype
The headline improvement comes from a materials swap. Majorana 1 used aluminum as its superconductor. Majorana 2 uses lead. That single change — plus switching the semiconductor layer to a combination of indium arsenide and indium arsenide antimonide — transforms qubit stability dramatically.
The numbers tell the story. Majorana 1 qubits held their state for roughly 20 milliseconds before decoherence set in. Majorana 2 achieves a mean qubit lifetime of 20 seconds, with some qubits holding coherence for a full minute. Operations run at the microsecond scale. Each qubit measures 1/100th of a millimeter. The topological gap protecting qubits from environmental noise is more than double the previous generation.
There was also a serious manufacturing challenge: lead is water-soluble. Standard chip fabrication washes materials in liquid processes. Microsoft had to solve the problem of using lead on a chip without it dissolving during manufacturing — a non-trivial feat that took years of materials engineering to crack.
AI Designed This Quantum Chip
Here is where the story gets recursive. Microsoft used its Discovery platform — an agentic AI system for scientific research — to design the materials that make Majorana 2 work. Discovery, which became generally available at Build 2026, deploys teams of AI agents to accelerate scientific discovery under human guidance.
So: Microsoft’s AI helped design the quantum hardware that may eventually power better AI. Whether that loop eventually closes is a question for 2029 and beyond. For now, Discovery is available as a free download for individuals with a GitHub Copilot account, and organizations can deploy their own Discovery agent teams on Azure.
The 2029 Target and What It Actually Means
Microsoft is now targeting a scalable, fault-tolerant quantum computer by 2029 — half the timeline it projected before. This sits within DARPA’s US2QC program (Underexplored Systems for Utility-Scale Quantum Computing), which gives the claim some institutional credibility. Read Tom’s Hardware’s full breakdown of the announcement for the technical specifics.
The roadmap runs through three phases: Foundational, Resilient, and Scale. “Practical” here means fault-tolerant computation on problems that classical computers cannot efficiently solve — drug development simulations, materials science, financial optimization at scale. It does not mean replacing your EC2 instances. Most developers will not need quantum skills for production workloads until well into the 2030s, if then.
The Physicists Are Not Convinced
This story requires a counterpoint, and it is a significant one. Microsoft’s topological qubit approach has faced sustained scientific skepticism since Majorana 1. Vincent Mourik, a physicist at the Helmholtz Research Centre in Jülich, stated plainly that the approach “is not going to work.” Researcher Andrew Legg contends that nothing in the presented data proves the existence of topological qubits or Majorana Zero Modes in these devices.
Microsoft was also forced to issue a correction to its original Majorana paper in 2025 after external scrutiny. Scientific American’s coverage details why physicists remain unconvinced. The pattern is consistent: Microsoft announces a breakthrough, the physics community pushes back, Microsoft refines its claims and tries again.
None of this proves Microsoft is wrong. Frontier science is messy and contested. But Majorana 2 has not settled the underlying scientific debate — it has extended it.
What Developers Should Do Right Now
The honest answer is: not much, urgently. But there are low-cost ways to stay positioned.
- Azure Quantum is available now. The platform supports Q#, Qiskit, and OpenQASM, runs on simulators without hardware access, and integrates with VS Code via the Quantum Development Kit.
- Q# in VS Code is the most practical starting point. Write and debug quantum programs against a local simulator — no waitlist, no cost, no physics PhD required.
- Microsoft Discovery app is free to download with a GitHub Copilot account — relevant if you work in scientific computing or R&D tooling.
- Watch the roadmap at quantum.microsoft.com. The milestones Microsoft hits (or misses) over the next 18 months will tell you more about whether 2029 is real than any announcement will.
IBM and Google both have competing approaches with hardware shipping today. Majorana 2 is Microsoft’s bet that topological qubits — fewer, more stable, inherently error-resistant — beat the brute-force qubit-count race. If the physics holds, it is a significant architectural advantage. If it does not, Microsoft will have spent a decade on a dead end. By 2029, we will know which story this is.
