Principal Investigator: Dr. Yue Zhao
As AI workloads continue to grow in scale and density, data center power requirements are rapidly approaching
the multi-megawatt range, making traditional low-voltage distribution systems increasing impractical. Conventional
architectures rely on heavy copper busbars, multiple conversion stages, and large rack-level power supplies, which
consume valuable space, increase losses, and limit power-density expansion. Transitioning to an 800 VDC backbone
addresses these challenges by reducing current, minimizing conductor size and losses, eliminating several conversion
steps, and freeing rack space, enabling seamless scaling from kilowatt-class racks to megawatt-class AI facilities.
Integrating Solid-State Transformers (SSTs) into this architecture provides additional benefits, including compact
high-efficiency AC/DC and DC/DC conversion, bidirectional power flow, voltage regulation, and improved power
quality, while also enabling direct interfacing with medium-voltage utility feeds for hybrid AC/DC operation. The
GRAPES project will perform a comprehensive assessment of 35 kV-class SSTs, examining their true value
proposition, impacts on the distribution grid, and performance at both component and system levels, in collaboration
with utilities and OEM partners, helping pave the way for next-generation, high-density AI infrastructure.
Recent Comments