UC San Diego to Test Solid-State Transformer for 2 MW AI Data Center Load
UC San Diego will test a medium-voltage solid-state transformer and flexible-load controls at the San Diego Supercomputer Center in an $8.48 million California Energy Commission-funded project designed to serve 2 MW of AI computing load while reducing power-conversion losses and supporting grid planning.
Published · 4 min read
UC San Diego announced on August 24 that its San Diego Supercomputer Center will become a test site for a medium-voltage solid-state transformer and flexible-load architecture intended to improve how high-density AI data centers connect to and interact with California’s power grid. The UC San Diego solid-state transformer demonstration is part of an $8.48 million California Energy Commission-funded project and matters because it will test the technology inside an operational AI computing facility rather than only in a laboratory environment.
The project is designed to serve 2 MW of AI computing load while targeting approximately 25% energy savings and a reduction of more than 50% in the footprint of power equipment. Researchers also plan to use operating data from the demonstration to develop models that California utilities and policymakers can use when evaluating large, flexible electricity loads.
The project would move power more directly from the grid to AI hardware
UC San Diego said in its August 24 announcement that the project will use a bidirectional solid-state transformer developed by San Diego-based Alderbuck Energy to convert medium-voltage alternating current directly to 800-volt direct current.
Conventional data centers typically move electricity through multiple conversion stages before it reaches computing hardware. Each conversion requires equipment, consumes space and introduces electrical losses that ultimately become heat that must also be managed by the facility.
The project is intended to test whether consolidating several of those stages into a solid-state transformer can create a shorter and more efficient path from the grid to high-density computing equipment.
Alderbuck Energy describes its Nexus Power Unit as a solid-state power-conversion platform intended for data centers, electric-vehicle charging and renewable-energy integration. For the UC San Diego demonstration, the architecture is designed around conversion from 12 kV AC to 800 VDC.
California is funding the demonstration with an $8.48 million grant
The California Energy Commission approved Agreement EPC-25-054 with the Regents of the University of California on behalf of UC San Diego for a grant of $8,484,515.
California Energy Commission business-meeting documents describe the project as a demonstration of a medium-voltage, grid-integrated solid-state transformer intended to enable accelerated and flexible interconnection of data centers using an 800-volt DC power architecture.
The funding also covers development of a flexible data-center load-capacity tool. That component is important because the project is not focused solely on improving efficiency inside the facility; researchers also want to determine how AI computing demand can become more responsive to grid conditions.
The demonstration will test whether AI loads can become grid resources
Large data centers are commonly evaluated by utilities according to the maximum electricity demand they may place on the grid. The UC San Diego project will examine whether some computing workloads can instead be shifted or managed without compromising the services the data center provides.
Emerald AI will contribute workload-management software to the project. UC San Diego says operating data will feed into a flexible-load capacity tool intended to give utilities validated models for planning around large data-center loads.
That approach could become increasingly important as utilities receive interconnection requests from AI facilities with load profiles that can be substantially larger and more dynamic than traditional commercial buildings.
The UC San Diego solid-state transformer project therefore has two related objectives: reduce the amount of energy and equipment required between the utility connection and computing hardware, and test whether the computing load itself can respond to power-system needs.
DERConnect will test the technology before live deployment
Before the new architecture is introduced into the operating supercomputer center, researchers plan to test it at DERConnect, UC San Diego’s grid research facility.
DERConnect can simulate large data-center loads and their effects on the grid while combining hardware-in-the-loop testing with real batteries, power electronics, controls and other distributed-energy equipment.
The facility includes five batteries totaling approximately 0.5 MW, grid emulation equipment, real-time simulation systems and thousands of controllable and simulated distributed-energy resources. Researchers can therefore expose new hardware and controls to realistic grid conditions before connecting them to an operating data center.
The project will operate inside UC San Diego’s campus microgrid
The demonstration also benefits from the unusual electrical environment surrounding the San Diego Supercomputer Center.
UC San Diego operates a 55 MW campus microgrid integrating solar generation, battery storage, combined heat and power, thermal storage and advanced grid controls. The university says the system supplies about 92% of the campus’s annual electricity needs and can island from the utility grid during outages.
That makes the campus a useful environment for evaluating how a large flexible computing load, advanced power electronics and distributed-energy resources can interact as part of a single power system.
The project is not claiming that solid-state transformers alone will solve data-center grid constraints. Interconnection capacity, generation availability, transmission infrastructure and local distribution upgrades remain separate constraints for large new loads.
What happens next
The solid-state transformer will first undergo simulations and hardware-in-the-loop validation at DERConnect before installation at the San Diego Supercomputer Center.
Once deployed, researchers plan to collect real operating data from heterogeneous AI research workloads and evaluate energy efficiency, equipment footprint, workload flexibility and grid interaction.
The most consequential result will be whether the project can validate its projected efficiency improvements and demonstrate enough repeatable load flexibility to influence how utilities model future AI data-center connections.
Jonas Muthoni is Editor-in-Chief of Microgrid Media, where he oversees reporting and analysis on microgrids, energy storage, distributed energy resources, data center power demand, grid modernization, resilience, renewable energy, and electricity policy and markets. His work focuses on the infrastructure, technologies, and regulatory developments reshaping the power system.
As Editor-in-Chief, Jonas leads Microgrid Media’s editorial strategy and standards, including story selection, source verification, technical accuracy, and the development of original reporting and analysis. His coverage draws on regulatory filings, government records, utility and company disclosures, technical documentation, and independent industry sources to explain significant developments across the evolving energy system.
