AI’s power boom is giving the transformer a high-tech makeover
The rapid growth of AI data centers is driving development of solid-state transformers that can manage high-voltage electricity more efficiently and flexibly than conventional equipment. Companies including DG Matrix, Nvidia, Google and Microsoft are developing or supporting architectures that could help data centers handle larger loads while reducing stress on the power grid.
AI data centers are increasing electricity demand and exposing limits in conventional grid equipment, prompting researchers and companies to develop solid-state transformers. A North Carolina State University team, with support from the New York Power Authority, tested a megawatt-rated prototype in Lenox, Massachusetts, while Raleigh-based DG Matrix is commercializing related equipment for data centers and other high-power customers.
Nvidia, Google and Microsoft published a draft specification for systems that convert medium-voltage grid power directly to 800-volt DC for AI data-center racks. The technology could reduce equipment footprints, improve efficiency and allow batteries and on-site generation to respond more quickly to grid conditions, although reliability, safety, certification and cost remain unresolved.
The article also describes the strain created by data centers, including a 3,800-megawatt shift on the PJM grid after a transmission fault in Ashburn, Virginia. Researchers and grid organizations say solid-state transformers and flexible load controls could help stabilize the system, but the technology requires substantially more field testing before widespread deployment.