Researchers at UC Merced are launching a $6 million project to use artificial intelligence to discover new materials that could cut the energy demands of AI computing in half. The initiative tackles a critical paradox: AI systems consume enormous amounts of electricity, yet AI itself might be the best tool to design the hardware that makes them more efficient. A standard ChatGPT query uses roughly 0.34 watt-hours of electricity, about 10 times more power than a Google search .

The energy problem is massive. U.S. data centers consumed 183 terawatt-hours of electricity in 2024, representing more than 4% of the country's total electricity consumption, roughly equivalent to the annual electricity demand of Pakistan . This growing appetite for computing power has prompted the University of California system to fund targeted research that could reshape how AI infrastructure operates.

What Are Topological Materials and Why Do They Matter?

The UC Merced team, led by chemical and materials engineering professor Elizabeth Nowadnick, is focusing on a special class of materials called topological materials. These materials have unusual electronic properties where electrons move in ways that differ fundamentally from conventional materials. The key advantage is that these materials can be switched on and off extremely rapidly while consuming minimal energy .

"We are investigating topological materials whose electronic structures can be rapidly switched with minor disturbances, meaning that the cost of each switching operation is minimized," explained Elizabeth Nowadnick, professor at UC Merced.

Elizabeth Nowadnick, Chemical and Materials Engineering Professor at UC Merced

The research team includes one postdoctoral researcher, Kuntal Talit, and two physics Ph.D. students, Haseeb Ahmad and Tharushi Ekanayake. They are using a computational tool called density functional theory (DFT) to simulate how topological materials would behave in computer systems before any physical testing occurs .

How Is AI Accelerating the Discovery Process?

The breakthrough lies in automating the discovery process itself. Rather than manually testing thousands of material candidates in laboratories, the team is building an AI system that can predict which materials are most promising before any physical experiments begin. This dramatically reduces the time and cost of materials discovery.

The researchers are collaborating with computer scientists at UC San Diego to develop an autonomous AI system called TritonDFT. This system automates density functional theory calculations and allows users to interact with the code using natural language, similar to how people use ChatGPT . The system orchestrates the entire workflow across multiple areas of expertise, including physics, computational chemistry, and high-performance computing.

• Natural Language Interface: Users can describe what they want to discover using everyday language instead of writing complex computer code, making the tool accessible to non-experts.
• Automated Workflow: TritonDFT handles the entire computational process automatically, coordinating between different specialized software tools and computing systems.
• Faster Results: By automating calculations that previously required manual setup and monitoring, researchers can perform simulations in a fraction of the time.

By identifying the most promising material candidates before laboratory testing, this AI approach could dramatically accelerate the discovery of new chip materials that switch faster while consuming less energy .

What Are the Two Main Goals of This Research?

The UC Merced team has set two specific objectives for the project. First, they aim to use AI to advance the discovery of topographical materials and develop a framework that can be applied to other functional materials like magnets, superconductors, and quantum materials. Second, they want to employ AI to develop new switching mechanisms for electronic devices, creating next-generation computing systems that operate faster while consuming significantly less power .

The project represents an unprecedented collaboration across the University of California system. The UC Office of the President issued a $6 million grant for the overall initiative, with UC Merced receiving $810,000 of that funding. The multidisciplinary effort involves principal investigators from five UC campuses: Santa Barbara, Merced, San Diego, Irvine, and Berkeley, as well as scientists from Lawrence Livermore and Los Alamos national laboratories .

"The rapid growth of artificial intelligence is accelerating both the opportunities for, and threats to, the United States' longstanding economic leadership. To help keep America in the lead, the University of California is scaling up its commitments in critical emerging areas of scientific research, such as AI, and moving forward with unprecedented speed to fund targeted research that fosters innovation," stated Theresa Maldonado, UC vice president of Research and Innovation.

Theresa Maldonado, UC Vice President of Research and Innovation

Why Does This Matter for the Future of Computing?

The implications extend far beyond academic research. If successful, materials that switch faster with less energy could fundamentally reduce the computational cost of running AI systems. This would lower electricity bills for data centers, reduce carbon emissions from computing infrastructure, and make advanced AI more accessible to organizations with limited resources. The research also demonstrates a broader principle: using AI to solve the problems that AI itself creates .

The project is part of a larger strategic initiative by the University of California to position American research at the forefront of AI innovation. As computing demands continue to grow exponentially, discovering materials that operate more efficiently becomes not just a scientific challenge but an economic and environmental imperative.