Introducing Majorana 2

While this line of materials research began long before the advent of agentic AI, the team used it to help manage the manufacturing of the new device, and Microsoft Discovery is being used more extensively for future Majorana materials work.

Critical parts of the Majorana quantum devices are designed atom by atom. To keep each atom in its correct spot, another material, an impurity, may be added to the crystalline structure. But adding too much or in the wrong way disturbs it, so it's a difficult balance to strike, said Zulfi Alam, corporate vice president for quantum at Microsoft.

"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," he said.

Agentic AI can analyze information at scale

The quantum computing project has many moving parts-software, architecture, design, the materials stack, fabrication processes, measurements and so on. A change in one area has ramifications that may require compensating elsewhere. AI agents help the team keep track of such complex, interrelated connections, Nayak said.

The quantum project also has huge quantities of data-nearly two decades' worth, in many different formats. Before AI, the data was stuck in silos. "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," Alam said.

In addition, the quantum team is spread across multiple countries, with very different specialties, such as physics, mechanical engineering and process engineering. It's impossible for any one person to be expert in everything. It's a common problem in interdisciplinary scientific research, which is why Microsoft's quantum team created an AI agent for organizing and analyzing information and making it easier for others to find.

"The AI is able to synthesize knowledge from all these different disciplines," Alam said, saving everyone the time and hassle of interviewing the specialists or of reading up on another subject. The agentic AI can "parallel process so much information in super short time to give you a recommendation," he said. The AI only offers guidance; it doesn't decide. "It's always 'scientist in the loop'."

Agentic AI can speed experiments

Creating a topological state requires setting hundreds of parameters. Then measurement, which is the key to performing quantum computations, can start. When done by a person, these processes each take weeks. In fact, measurement is so difficult and time consuming that the team had tried to automate it a few years ago using earlier forms of machine learning, but it wasn't possible, Alam said.

Using agentic capabilities available in Microsoft Discovery, the team was able to create an AI agent specialized for this job, which cut the cycle time by orders of magnitude, he said.

AI's pattern-recognition abilities helped with the difficult task of measuring what state the qubit is in and detecting whether there's an even or odd number of billions of electrons on a semiconductor wire. AI agents run the process automatically and continuously, building a 3D map of the conditions that a single scientist would never be able to do in the same way, Alam said.

"Using agentic AI to automate the measurements was a game changer," he said. "It goes through some math and starts saying, 'Hey, where do I find the lowest point where everything sort of works?' And it can do all these voltage adjustments in parallel, which a human cannot do. The way our minds work, we are more linear."

Agentic AI can quiet the noise

Data isn't information-it needs to be filtered, analyzed and put into context to have meaning. For example, the team developed an AI agent that was able to combine physics, device and institutional knowledge to filter raw data from the quantum team's fabrication process and sniff out an uncalibrated temperature sensor reading that was throwing things off.

Alam compares the process to the AI summary of a Teams call, which skips over friendly banter to list the three or four key points. "That's exactly what the AI is doing on a grander scale when science is involved," he said.

Microsoft Discovery was built as a platform that pairs AI with the scientific method, and many of the agentic AI tools that the quantum team is using are transferable and relevant to scientific exploration in other domains.

This fundamentally new type of Frontier R&D lets a scientist "be the anchor point and look at many, many different disciplines all at the same time with a very high fidelity and be able to draw correlations from that," Alam said. "It is the essence of what every single high-performance, cutting-edge team wants to do."

Related links:

Learn more: Microsoft's Majorana 2 is here

Read more: Majorana 2 - Microsoft's scalable quantum processor with reliable, long-lasting qubits

Read more: 20 Second Parity Lifetime in an InAs-Pb Device

Read more: Announcing Microsoft Discovery general availability for R&D and Microsoft Discovery app preview

Download: Microsoft Discovery app

Learn more: Microsoft Discovery

Read more: Microsoft's Majorana 1 chip carves new path for quantum computing

Lead photo: Majorana 2, a next-generation quantum chip built with the help of Microsoft Discovery's agentic AI. Photo by John Brecher for Microsoft.

Catherine Bolgar writes about AI and innovation at Microsoft, from advances in quantum computing to how AI is helping ordinary people. Previously, Catherine wrote about technology and business for a number of publications, and she was an editor at the Wall Street Journal in New York and Brussels. She taught high school math in Kenya, where she learned Swahili. She currently lives in France. You can contact Catherine on LinkedIn .