Data center AI boom drives solid-state transformer investment

The conventional power transformers that the US grid relies on are painstakingly assembled by hand, with a design dating back to the 1880s. Because the largest ones are custom-built for each utility substation, customers now face waits of up to several years for delivery, which delays both grid expansion and replacement of aging equipment. Solid-state transformers, by contrast, use high-frequency semiconductor switching and can be mass-produced, offering a potential path around those manufacturing bottlenecks.

Tech companies building large AI data centers are a key early adopter because these facilities are moving to DC power architectures to support energy-hungry AI chips. A solid-state transformer can perform the AC-to-DC conversion in one step, which professor Srdjan Lukic described as a "one magic box" that eliminates infrastructure and interoperability challenges. This adoption is seen as a way to "derisk" the technology for broader use, including EV charging and eventually home DC distribution.

The NC State demonstration at EPRI's lab in Lenox, Massachusetts, is the culmination of a collaboration with the New York Power Authority that began in 2018. The success of this field testing, along with continued commercialization efforts, will likely determine how quickly the technology moves beyond data centers into wider grid applications.