UTSA researchers to tackle AI’s growing energy footprint

MARCH 14, 2025 — Many groundbreaking inventions are energy-intensive in their infancy, and AI is no exception. A multidisciplinary team of researchers from UTSA and four partnering institutions will pursue fundamental research in a project dubbed ESTEEM to improve the energy efficiency of microelectronics and AI with funding from the U.S. Department of Energy (DOE). Dhireesha Kudithipudi, founding director of MATRIX: The UTSA AI Consortium for Human Well-Being and the Neuromorphic AI (NuAI) Lab, will be a co-PI on the project.

AI is significantly more energy-intensive than traditional computing models. It is estimated that one day of ChatGPT’s energy use could run a U.S. home for 46.5 years. Last year, NPR reported that a single query to ChatGPT uses approximately as much electricity as a light bulb consumes in 20 minutes.

Through the project, titled Enabling Science for Transformative Energy-Efficient Microelectronics (ESTEEM), Kudithipudi and her research collaborators will strive to curb this energy use. Over the next four years, they will take an “atoms to algorithms” approach, exploring ways to make AI hardware (microelectronic chips) and software more energy efficient.

“As AI models become increasingly complex, traditional computing architectures are reaching their limits,” Kudithipudi said. “Hardware-software co-design is essential for developing real-time, high-performance AI systems that are both efficient and scalable.”

“This work will lead to more energy-efficient systems that will help unleash the power of AI into all aspects of our lives.”

To innovate at the atomic level, one part of the project team will test new or improved materials, surface processing and control as well as the chemistry, synthesis and fabrication of microelectronics. Meanwhile, another group will aim to develop algorithms that consume less energy while producing the same results.

In the final component of the project, researchers from all five collaborating institutions will take a co-designed approach to ensure that both the software and hardware work seamlessly together.

Kudithipudi specializes in developing AI algorithms that can quickly learn and adapt to new situations, similar to how the brain works. These algorithms are designed to work with neuromorphic hardware, which mimics the structure and function of the human brain.

She is especially interested in whether new AI models inspired by the human brain can learn new tasks without interfering with previously learned tasks.

“Dr. Kudithipudi and the team at MATRIX AI are leaders in the area of neuromorphic computing, and this work will lead to more energy-efficient systems that will help unleash the power of AI into all aspects of our lives,” said Eric Brey, dean of the Margie and Bill Klesse College of Engineering and Integrated Design at UTSA. “This project aligns with the overarching commitment of our Klesse College faculty to providing solutions that enable our society to thrive.”

ESTEEM will be housed in the National Accelerator Laboratory (SLAC), a collaboration between Stanford University and the DOE, and led by Paul McIntyre, SLAC’s associate director. UTSA’s NUAI Lab will provide additional equipment and support, as will partnering institutions Georgia Tech, Northwestern and the University of Tennessee, Knoxville.

UTSA students will have the opportunity to participate in the project and design new types of AI algorithms.

Kudithipudi and her collaborators will receive $12 million in funding to advance their work. ESTEEM is part of MEERCAT — one of three microelectronics science research centers that received $179 million in combined funding from the DOE.

About MATRIX AI

Bringing together over 80 scientists and AI experts from UTSA, UT Health San Antonio, the Southwest Research Institute and the broader San Antonio region, MATRIX AI leads the way in pioneering transdisciplinary AI solutions.

Focused on four key research thrusts — machine learning and deployment, augmenting human capabilities, neuro-inspired AI and trustworthy AI — and strategically partnering with industry leaders and federal research labs, MATRIX AI is tackling significant AI challenges while addressing the nation’s growing need for interdisciplinary AI research talent. With multiple dedicated research centers and a plethora of collaborative projects, MATRIX AI is the primary hub for advancing AI research at UTSA and in San Antonio.

— Audrey Gray