A new $8.48 million partnership could change how AI data centers receive electricity while helping prevent them from overwhelming California‘s power grid
Story by:
- Kimberly Mann Bruch
–
kbruch@ucsd.edu
Published Date
Story by:
- Kimberly Mann Bruch
–
kbruch@ucsd.edu
Article Content
- San Diego Supercomputer Center will test a new solid-state transformer that reduces energy waste, lowers costs and shrinks the footprint of power equipment without performance loss.
- The project is designed to help data centers support grid reliability while providing valuable data for utility planners and policymakers.
- The project will create career pathways that connect residents of under-resourced communities with jobs in data center and energy infrastructure operations.
The San Diego Supercomputer Center (SDSC) at the University of California San Diego Halıcıoğlu School of Data Science and Computing will serve as the testing ground for a new $8.48 million California Energy Commission (CEC) funded project that could transform how AI data centers receive and manage electricity.
The project will make SDSC one of the first operational AI data centers in the United States to demonstrate a new power architecture designed to reduce energy waste, lower costs, and improve grid reliability. Beyond the technology itself, the initiative also includes what the project team describes as one of California’s first equity-centered data center workforce development efforts, creating pathways into careers in the rapidly growing AI and energy infrastructure sectors.
Meeting the Growing Energy Demands of AI
As artificial intelligence continues to expand, so does the energy required to support it. Computers powering AI have become dramatically more energy intensive. A single rack of AI hardware can require up to 50 times more electricity than a conventional server rack installed just a few years ago.
The challenge is not only how much power these systems consume, but also how that power is delivered. Traditional data centers rely on multiple conversion stages before electricity reaches computing equipment, with each step wasting some energy as heat.
To address this problem, the UC San Diego team is working with industry leaders to develop a power delivery system that eliminates many of those conversion stages. The SDSC-based project will provide a testbed for electricity to move more directly and efficiently from the grid to the computing equipment.
“This research has the potential to fundamentally transform how AI data centers interact with the electric grid,” said Chancellor Pradeep K. Khosla. “By cutting costs and shrinking our environmental footprint, it paves the way for sustainably scaling this critical technology.”
A Smarter Path from Grid to Rack
At the heart of the project is San Diego-based Alderbuck Energy, whose team develops power conversion and intelligent energy management solutions for data centers and other large-load customers. Alderbuck’s bidirectional solid-state transformer (SST), which converts medium-voltage alternating current directly into 800-volt direct current through a single compact device, will play a big role in the new CEC-funded project. By simplifying the power delivery chain, the team expects the system to improve efficiency, reduce equipment requirements and lower installation and operating costs. The project aims to serve two megawatts of AI computing load while reducing the footprint of power equipment by more than 50% and achieving projected energy savings of approximately 25%.
“Data centers are no longer just consumers of electricity, as they can be active participants in grid stability,” said Brian Balderston, director of infrastructure and data centers for SDSC’s Research Data Services Division.
But the case for this architecture starts with efficiency: Every unnecessary conversion stage between the grid and the computation is energy lost, and at AI scale, that adds up fast. “This project shows us what a better path looks like in practice and will produce the real-world evidence that moves this architecture from promising to proven,” he said.
Balderston said the project goes beyond hardware by combining Alderbuck’s grid management controller with computer orchestration software by Emerald AI, which is a flexibility management platform for AI infrastructure. Together, they allow the system to maintain reliable power inside the data center, while also responding to conditions on the broader grid, turning the data center from a passive load into an active grid resource.
Emerald AI’s role is to make AI data centers grid-responsive, transforming them into valuable grid partners. Its platform, Emerald Conductor, acts as the intelligent interface between a facility’s compute system and its power infrastructure dynamically shaping how much electricity AI jobs draw by slowing, shifting, or briefly pausing batchable workloads, while protecting the performance of jobs that cannot tolerate delay.
This approach was first validated in a peer-reviewed Nature Energy study, in which Emerald AI, NVIDIA, Oracle, Salt River Project and the Electric Power Research Institute successfully reduced the power draw of a live AI cluster in Phoenix by 25% for three hours without breaching performance commitments. That trial was the first of five live demonstrations Emerald AI has completed at commercial data centers across Arizona, Illinois, Virginia, Oregon and the United Kingdom. This includes a recent London test at Nebius’s new AI Factory, where Emerald AI and National Grid demonstrated that high-performance AI infrastructure could cut electricity demand by up to 40% without any performance loss. Emerald AI’s software is now being integrated with NVIDIA’s DSX Flex stack to establish the benchmark architecture for power-flexible AI factories.
The SDSC facility’s diverse mix of scientific and AI research workloads – running inside the UC San Diego campus microgrid with its own distributed energy reoad flexibility data it produces will feed directly into UC San Diego’s new flexible load capacity tool, providing validated models to help California policymakers and utilities plan for large DC loads
“Most interconnection planning still treats a data center’s peak demand as a fixed number the grid either serves or doesn’t,” said Ayse Coskun, chief scientist at Emerald AI and a professor at Boston University. “Across a series of real-world demonstrations, we’ve shown that assumption is wrong. A meaningful share of AI computing has flexibility that can be used to actively support the grid.”
“What makes the UC San Diego-led project different is the combination. “We’re integrating Emerald AI’s software with Alderbuck’s new solid-state transformer technology at a facility running real heterogeneous research workloads, all inside a campus microgrid,” Coskun continued. “That’s a much broader test, and it’s the one that will produce the validated evidence and models a California utility planner can use for the new flexible load capacity tool.”
Building Tools for California’s Energy Future
In addition to testing the new hardware, the UC San Diego team will develop a flexible load-capacity tool that utilities and policymakers can use to better understand how large electricity users, including data centers and EV fast-charging stations can be integrated into the grid.
As part of this innovation, San Diego Gas and Electric Company (SDG&E) will serve as the utility technical advisor and grid integration partner by providing expertise in utility operations, grid integration, communications standards and evaluation of data center load characteristics. UC San Diego will lead the development, testing and assessment of technologies intended to improve utility visibility into and integration of grid-interactive data centers with SDG&E providing utility perspective and feedback throughout the effort, sharing relevant forecasting and operational insights.
“Meeting California’s growing energy needs will require collaboration across utilities, technology providers and research institutions,” said SDG&E Senior Vice President and Chief Commercial Officer Miguel Romero. “Building upon our existing relationships, this project creates an opportunity to evaluate innovative approaches that could help large energy users operate more efficiently, while providing valuable insights for future grid planning, reliability and long-term affordability for customers.”
Through the project, SDG&E will provide guidance on typical load profiles and advise on how these profiles can be represented in modeling tools. This work will provide valuable insights into the integration of emerging data center technologies while supporting reliable grid operations.
“California is positioned to lead the next wave of AI infrastructure, both for large, centralized data center campuses and for more distributed facilities,” said Alderbuck Energy Chief Strategy Officer Kimberly McGrath. “This deployment is designed to show how medium-voltage-to-DC power infrastructure can help operators bring high-density compute capacity online faster, while maintaining the reliability data centers require.”
Flexible Data Center project overview demonstrating integration of solid-state transformers, high voltage power supply within the San Diego Supercomputer Center, Emerald AI and distributed energy re
A First-of-Its-Kind Demonstration
The project could become the first field validation of a medium-voltage-connected solid-state transformer at an operational California data center. Before deployment at SDSC, the technology will undergo extensive testing including simulations at DERConnect, UC San Diego’s National Science Foundation-funded hardware-in-the-loop testbed. According to DERConnect Director Jan Kleissl, researchers will model AI workloads and grid interactions before introducing the system into a live operating environment. Kleissl, who is the Henry G. Booker Endowed Chair and a professor at the UC San Diego Jacobs School of Engineering Department of Mechanical and Aerospace Engineering, also serves as principal investigator on the newly funded CEC award, alongside SDSC’s Balderston and Mike Ferry, who is director of UC San Diego’s Energy Storage Group.
“Building on over a decade of experience deploying and studying battery energy storage systems, EV charging infrastructure, grid-interactive building controls and DER (distributed energy res microgrid, we are uniquely positioned as an operational proving ground for next generation technologies that will reshape and modernize our grids,” Ferry said
Workforce and Community Impact
The project’s impact extends beyond technology and grid modernization. Working with the Good For Others Foundation, a community-based organization focused on workforce partnerships and equitable career pathways, the team will develop a career pathway program that has been designed to connect residents of disadvantaged communities with jobs in high-power, grid-interactive data center operations.
“We are excited to partner on this project because it connects cutting-edge energy innovation with real workforce pathways for communities that have historically been left out of these opportunities,” said Good For Others Foundation President and CEO John Valencia. “We see this as a chance to help build a stronger, more equitable clean energy economy while creating durable careers in a fast-growing industry.”
“We are coordinating a multidisciplinary team that brings together expertise in grid planning, power electronics, AI computing and utility-scale data center operations,” said SDSC Director Frank Würthwein. “The results of this project will help shape future interconnection standards, grid planning practices and deployment strategies for the next generation of AI infrastructure, EV charging systems and other large electrical loads across California and beyond.”
Learn more about research and education at UC San Diego in: Artificial Intelligence
Read more news about: Halıcıoğlu School of Data Science and Computing,
Artificial Intelligence
