UC Santa Barbara Researchers Address Growing Energy Needs of AI and Quantum Technology

UC Santa Barbara researchers are addressing the escalating energy requirements of advanced computing, specifically artificial intelligence (AI) and quantum technologies. The university’s Institute for Energy Efficiency (IEE) has provided seed funding for five distinct projects to develop more energy-efficient solutions for these growing computational demands.

Steven DenBaars, director of IEE, emphasized the increasing urgency of challenges posed by the energy demands of AI, data centers, and the next generation of quantum technologies. He noted that these awards allow researchers to tackle problems early, test ambitious ideas, and advance promising technologies toward real-world application. The projects were selected through IEE’s 2026–27 Research Seed, Software Impact, Gauthier Quantum, and Photonics Seed grant programs, which support early-stage work designed to open new research directions and establish groundwork for larger investments. DenBaars, a Distinguished Professor of Materials and Electrical and Computer Engineering, stated that the goal is to give researchers the freedom to pursue ideas that are "high-risk, high-impact" and capable of opening entirely new directions.

One project, led by Xin Eric Wang, an assistant professor of computer science, is focused on helping AI agents complete tasks using fewer “tokens” and less computation. As AI agents undertake longer tasks—such as conducting research, navigating software, or completing digital workflows—they generate and reprocess vast numbers of tokens, which are units used to reason and respond. The computational cost increases with every step. Wang’s team is developing an open-source system that trains models to achieve successful answers through shorter reasoning paths and reduces the amount of information long-running agents must repeatedly process. The team aims for a fraction of current token usage while maintaining performance. This work builds on earlier research from Wang’s group that enables a model to estimate in advance how costly different routes to a solution are likely to be.

Another initiative involves electrical and computer engineering professors Mahnoosh Alizadeh and Ramtin Pedarsani, who are developing AI systems that allocate computation based on the difficulty of each request. They are focused on creating "energy-proportional AI," which uses lighter configurations for simpler requests and reserves the full model for more demanding ones. Their goal is to create AI systems that consume only as much computing power as each task requires. An online learning system would choose the lowest-energy option that can still meet a required level of quality. The researchers plan to measure the energy required to produce a correct response, rather than solely relying on model size or the number of computations performed. This work builds on prior research by their team that identified substantial redundancy within large vision-language models, demonstrating that significant layers could be removed while retaining performance.

Sriram Krishnamoorthy, a professor in the Materials Department, and Galan Moody, a professor of electrical and computer engineering, are exploring whether semiconductor-based quantum systems can operate at or near room temperature. Many quantum technologies currently demand extreme operating conditions, including cryogenic temperatures and ultra-high vacuum. Their project focuses on “color centers,” which are atomic-scale defects in crystalline materials that can interact with light and, in some cases, emit individual photons capable of carrying quantum information. They are specifically investigating these defects in gallium oxide, an ultra-wide-bandgap semiconductor that could support stable single-photon emission at room temperature. The collaboration brings together Krishnamoorthy’s expertise in high-quality gallium oxide materials with Moody’s expertise in quantum photonics. Their objective is to demonstrate single-photon emission from gallium oxide for the first time, potentially opening a path toward compact quantum light sources that require significantly less energy-intensive infrastructure.

Professors of mechanical engineering Yangying Zhu and Igor Mezić are working to reduce the substantial energy burden of thermal management in data centers, which can account for 30% to 50% of total energy use. They are developing a more adaptive system to manage heat by directing cooling resources to the processors that need them most as computing loads change. They stated that this approach enables significant reductions in overall cooling energy consumption by reducing overcooling of low-load devices. Their system integrates advanced chip-level cooling with an AI-enabled control system capable of redistributing coolant across computational processing units and graphics processing units in real time. The researchers estimate that a successful demonstration could cut cooling energy use in AI data centers by 20%. At the chip level, tiny silicon channels would carry coolant close to heat-generating electronics, including systems where liquid absorbs heat by turning to vapor. The larger cooling network would use predictive models, based on Koopman theory, to anticipate shifting thermal loads and adjust coolant flow accordingly, ensuring a system that responds as computing demands shift, rather than spending energy cooling every component equally.

Dmitri Strukov, a professor of electrical and computer engineering, is exploring specialized hardware to make probabilistic computing tasks faster and more energy-efficient than conventional processors. These tasks, which involve generating random numbers and repeated mathematical operations known as dot products, are crucial across various fields, from artificial intelligence and physics to statistics and cryptography, but can become major computational bottlenecks. Strukov explained that many modern workloads are "fundamentally ill-suited to general-purpose processors" and even highly data-parallel architectures, such as graphics processing units. He noted that highly efficient, purpose-built computational circuits continue to emerge, offering dramatic gains in performance and energy efficiency. This project reflects a broader shift toward designing specialized hardware for computationally demanding tasks that conventional processors handle inefficiently. Strukov and his collaborators will use this work to gather preliminary data for a scalable framework for probabilistic computing and to support a larger external funding effort.