Microsoft's quantum push now includes AI-built chips: Majorana 2 explained

Built with the help of AI agents, Microsoft's Majorana 2 takes a different approach to quantum computing, focusing on hardware-level stability as rivals like Google push scaling and error correction

Microsoft has unveiled Majorana 2, its next-generation quantum chip, marking a significant step in its long-running effort to build a scalable quantum computer. Unlike conventional chip announcements, however, this one is as much about how the chip was built as what it can do.

 

According to Microsoft’s announcement, Majorana 2 is a topological quantum chip developed with the help of its Microsoft Discovery platform, which uses agentic AI systems to accelerate scientific research. The company says the chip delivers a 1,000-fold improvement in qubit reliability over its previous generation, while also cutting its roadmap to a scalable quantum computer to 2029.

 

What quantum chips are and why they matter

Quantum chips are fundamentally different from traditional processors. While classical computers use bits that represent either 0 or 1, quantum computers rely on quantum bits, or qubits, which can exist in multiple states at the same time through a phenomenon known as superposition (the ability to represent both 0 and 1 simultaneously). They can also be linked through entanglement, where the state of one qubit is directly related to another, allowing them to process complex combinations of data in parallel.

This allows quantum systems to tackle problems that are effectively impossible for classical machines, including molecular simulation, advanced materials design, and optimisation problems across industries such as energy, healthcare and finance.

 

However, quantum systems are also extremely fragile. Qubits are highly sensitive to external disturbances like heat, radiation, or electromagnetic interference, which leads to errors and limits their practical use. Much of the industry’s progress has therefore focused on improving stability and error correction, rather than just increasing raw compute power.

What makes Majorana 2 different

Microsoft’s approach to quantum computing differs from many of its peers because it focuses on topological qubits, which aim to improve stability by encoding information in the structure of the system itself, instead of relying on delicate quantum states that are prone to errors.

 

Majorana 2 builds on this by introducing a new materials stack, replacing the aluminium-based structure used in earlier designs with a lead-based superconductor (a material that can conduct electricity with zero resistance under specific conditions). According to Microsoft, this change significantly improves the chip’s ability to shield qubits from external interference.

 

The result is a significant improvement in reliability. Microsoft says the qubits in Majorana 2 can maintain their quantum state 1,000 times longer than the previous generation, with a mean lifetime of around 20 seconds and some lasting as long as one minute.

 

This is a substantial shift when compared to many quantum systems, where qubit lifetimes are typically measured in microseconds. Microsoft compares the improvement to extending a smartphone battery from a single day to nearly three years on a single charge.

 

The chip also combines this reliability with fast operation speeds and smaller qubit sizes, which the company says puts it on track toward building a commercially viable quantum system within the next few years.

How AI helped build the chip

A key part of the Majorana 2 story is the role of Microsoft Discovery, the company’s agentic AI platform for scientific research.

 

Rather than being used only for simulations or modelling, these AI agents were integrated into the development process itself. According to Microsoft, the quantum team used AI to manage workflows, automate measurements, optimise fabrication processes (the process of physically building the chip), identify hidden flaws, and propose new design approaches.

Quantum hardware development involves a large number of interconnected variables, from materials and fabrication to measurement and system design. Changes in one area can affect multiple others, making it difficult for researchers to track all dependencies manually.

 

Microsoft says its AI agents were able to process large volumes of experimental and historical data, identify patterns that humans might miss, and suggest more efficient paths forward.

 

In some cases, AI reduced the time required for experiments from weeks to significantly shorter cycles by automating measurement processes and exploring multiple configurations in parallel.

 

At the same time, the company emphasises that these systems operate with a “scientist in the loop” model, where AI provides recommendations but final decisions remain with researchers.

How it compares to Google’s Willow chip

In December 2024, Google introduced its Willow quantum chip, which demonstrated significant progress in error correction and system performance. According to Google, Willow was able to reduce errors as the number of qubits increased, achieving a key milestone known as operating “below threshold” (a point where adding more qubits actually improves reliability rather than increasing errors).

 

The company also reported that Willow completed a benchmark computation in under five minutes that would take one of the world’s fastest supercomputers an estimated 10 septillion years to perform.

 

While Google’s approach is based on superconducting qubits and focuses heavily on scaling systems and reducing errors through correction techniques, Microsoft is taking a different route with topological qubits, which aim to improve stability at the hardware level itself.

 

In Google’s case, qubits are inherently fragile and prone to errors, so the focus is on building systems that can detect and correct those errors as the number of qubits increases. Microsoft, on the other hand, is attempting to reduce the need for such correction by designing qubits that are more stable by default, encoding information in a way that is less affected by external disturbances.

 

Both approaches are aimed at solving the same core challenge: making quantum systems reliable enough for real-world use.

What this means

Majorana 2 does not represent a finished quantum computer, but it signals progress toward one. The key takeaway is not just the improvement in qubit stability, but the method behind it. By combining new materials science with agentic AI-driven research, Microsoft is attempting to accelerate a field that has historically progressed slowly due to its complexity.

 

If the company’s timeline holds, a scalable quantum system could arrive within the next few years, potentially unlocking new capabilities in scientific research, industrial design, and large-scale optimisation.