Navitas Surges 200% Following Strategic Partnership with NVIDIA on 800V AI Power Infrastructure

On May 21, Navitas Semiconductor announced a strategic collaboration with NVIDIA to co-develop a next-generation 800V high-voltage direct current (HVDC) power architecture, marking a significant step in redefining the energy infrastructure for AI data centers. The new “Kyber” rack-level system, supporting GPUs including Rubin Ultra, will leverage Navitas' gallium nitride (GaN) and silicon carbide (SiC) technologies as core components.

According to media reports, this development signals a technological leap in the data center industry, especially in handling gigawatt-level AI computing demands. Navitas’ advanced materials are expected to enhance energy efficiency and reduce reliance on copper. Following the announcement, Navitas’ share price surged by as much as 200% in post-market U.S. trading on Wednesday.

Current data centers rely on 54V rack-level distribution systems, which are limited to several hundred kilowatts of power delivery. These systems require heavy copper busbars to transmit low-voltage power across racks. As power requirements exceed 200kW, issues related to power density, copper usage, and system efficiency become critical.

NVIDIA’s proposed architecture aims to address these challenges by converting 13.8kV AC grid power directly to 800V HVDC at the perimeter using solid-state transformers (SSTs) and industrial-grade rectifiers. This design reduces the need for multiple conversion steps and improves overall reliability.

By increasing voltage and reducing current—as per the I²R loss principle—copper usage can be significantly reduced. The shift to 800V HVDC allows copper wire thickness to be lowered by up to 45%, whereas traditional 54V systems require over 200kg of copper per megawatt, posing sustainability challenges for next-gen AI centers.