Powering AI's Future: A New Grid Architecture Goes Live in California

SAN DIEGO, CA – June 23, 2026 – In a move that signals a critical shift in how we power the artificial intelligence revolution, Alderbuck Energy has announced the deployment of its cutting-edge solid-state transformer platform at the San Diego Supercomputer Center (SDSC). This isn't just another hardware installation; it's a real-world test of a new electrical backbone designed to solve the single biggest bottleneck for AI's growth: power. The project, backed by the California Energy Commission, aims to validate a novel 12 kV AC-to-800 VDC power architecture, potentially rewriting the rules for data center construction, efficiency, and grid interaction.

For leaders navigating the AI boom, the promises of faster model training and revolutionary applications are often overshadowed by a more pragmatic and pressing concern. The infrastructure to support these ambitions is straining at the seams, creating a chasm between digital demand and physical reality. This deployment at one of the nation's premier supercomputing hubs offers a grounded, data-driven look at a potential solution.

The Unseen Crisis Fueling AI

The explosive growth of generative AI has created an insatiable appetite for electricity. Traditional data centers, designed for rack power densities of 5-10 kilowatts, are ill-equipped to handle the new reality. AI workloads, driven by clusters of power-hungry GPUs, are pushing rack densities to 30 kW, 50 kW, and in some cases, over 100 kW. This dramatic escalation in power density has created a cascade of challenges that threaten to stall AI development.

The most significant hurdle is the grid itself. Utilities and grid operators are struggling to accommodate requests for massive, concentrated power loads. Interconnection queues in major data center markets are backlogged for years, making "time to power" the primary constraint on growth. Inside the data center, legacy 480 V AC power distribution systems, with their multiple, inefficient conversion steps, are proving to be a costly bottleneck. Each conversion from AC to DC and back again wastes energy as heat, which in turn drives up cooling costs—a vicious cycle that inflates operational expenses and the facility's carbon footprint.

Industry analyses project that global data center electricity consumption could double by 2030, largely driven by AI. The problem is so acute that data center developers are no longer just looking for real estate; they are on a global hunt for available megawatts. The physical footprint of traditional power equipment—bulky transformers, switchgear, and uninterruptible power supplies (UPS)—also consumes valuable floor space that could otherwise be used for revenue-generating servers. This project at SDSC tackles these pain points head-on.

Reinventing Power from the Grid to the Chip

The solution being tested is a fundamental rethinking of power delivery, centered on two key technologies: solid-state transformers (SSTs) and 800-volt direct current (VDC) distribution. Alderbuck Energy's Nexus Power Unit™ (NPU) is an SST, a power-electronics-based device that replaces conventional, hulking copper-and-steel transformers with a compact, intelligent, and highly efficient system.

Unlike traditional transformers that are locked into a fixed conversion ratio, SSTs use high-frequency semiconductor switching to digitally manage power flow. This allows them to directly convert medium-voltage AC power from the utility (in this case, 12 kV) straight to 800 VDC, the high-voltage direct current architecture championed by AI leaders like NVIDIA and endorsed by the Open Compute Project. This direct path eliminates multiple intermediate conversion stages, which is the source of the project's ambitious goals: a 4 percent improvement in energy efficiency, a more than 50 percent reduction in the power equipment footprint, and a more than 50 percent faster installation timeline.

For data center operators, these numbers represent a paradigm shift. A smaller footprint means more white space for servers. Faster installation means bringing capacity online months sooner. And higher efficiency translates directly to lower operating costs and a more sustainable facility. As Brian Balderston, Director of Infrastructure and Data Centers for SDSC, noted, "the shift to 800 VDC will be one of the most significant architectural changes we'll see in the years ahead. AI is fundamentally changing what data centers demand from the grid, and efficiency gains at scale are why this architecture matters and why we need to validate it in real operating conditions."

De-Risking the Future at UC San Diego

Moving a data center—especially a high-performance computing facility like SDSC—to a new power architecture is not a trivial undertaking. This is where the project's collaborative, research-driven approach provides crucial assurance. Before the NPU is installed in the live data center environment, it will undergo rigorous testing at DERConnect, UC San Diego’s state-of-the-art utility-scale testing facility.

Researchers will use a technique called hardware-in-the-loop simulation, essentially creating a digital twin of the data center's AI workloads and the local utility grid. This allows them to subject the SST to a full range of operating conditions, fault scenarios, and dynamic load changes in a controlled environment, effectively de-risking the system before it ever touches the critical compute infrastructure.

"UC San Diego’s campus microgrid and DERConnect facility provide a unique environment to test how high-density data center loads can interact with the grid in real operating conditions,” said Mike Ferry, UC San Diego Energy Storage Group Director. This validation is critical for building confidence among operators and utilities who are traditionally risk-averse. The partnership, which includes UC San Diego as the prime contractor, SDSC, Alderbuck, San Diego Gas & Electric Company, and software provider EmeraldAI, exemplifies the public-private collaboration needed to accelerate such foundational innovations.

From Power Drain to Grid Partner

Perhaps the most far-reaching implication of this technology lies beyond the data center walls. The intelligence and bidirectional capability of solid-state transformers position data centers to evolve from being massive, passive consumers of electricity into active, responsive partners with the grid.

The NPU can precisely control power quality, mitigate harmonics, and even feed power back into the grid from onsite energy storage or generation. This flexibility is invaluable to utilities struggling to balance intermittent renewable energy sources like solar and wind. A data center that can intelligently modulate its power consumption in response to grid signals becomes an asset, not just a liability.

Recognizing this, a key deliverable of the project is a "flexible load capacity tool" designed to help California policymakers and utilities plan for the integration of large DC loads. This project isn't just about solving a problem for data centers; it's about creating a blueprint for how the state can accommodate the massive power needs of AI, EV fast charging, and industrial electrification while advancing its clean energy goals.

“California is positioned to lead the next wave of AI infrastructure, both for large, centralized data center campuses and for more distributed facilities,” said Kimberly McGrath, Chief Strategy Officer at Alderbuck Energy. The deployment at SDSC is designed to demonstrate that this new infrastructure can deliver the speed and capacity AI demands, while also providing the reliability and grid-friendliness that utilities require. The results from this trial will be watched closely, as they may well illuminate the path forward for powering the next frontier of computing.