en.Wedoany.com Reported - Recently, German Infineon Technologies announced its entry into the NVIDIA MGX AI Factory ecosystem, providing power management solutions for next-generation AI data center server racks. The collaboration revolves around the NVIDIA MGX architecture and 800V DC power supply system, aiming to improve power efficiency and power density from the grid to processor cores in the context of continuously expanding AI model scales and rising computing density.
The focus of this collaboration is on upgrading the power supply architecture of AI server racks. Infineon disclosed that its power management solution will support the NVIDIA MGX architecture and 800V DC power supply architecture, an open, modular reference architecture for AI factories. The 800V DC MGX-compatible power rack can help existing AI infrastructure improve computing performance and power density, and provide an upgrade path for future higher-density AI infrastructure. As the scale of AI training and inference tasks expands, data centers need to accommodate more accelerated computing resources under limited physical space, power access, and cooling conditions. The efficiency loss, conversion stages, thermal management, and protection capabilities of the power supply chain are becoming key constraints for computing capacity expansion.
Infineon's technical approach covers the complete power conversion chain from the grid to core voltage, involving semiconductor materials such as silicon, silicon carbide, and gallium nitride. The company stated that gallium nitride technology can be used in ultra-compact bus converters at switching frequencies close to 1MHz, while silicon carbide JFET technology and dedicated control ICs are adapted for the protection and hot-swap functions of native 800V server boards. The relevant power management solution can convert 800V voltage to 50V, 12V, and even further down to 6V, bringing DC power closer to the load end inside the rack.
AI data centers have historically focused more on GPUs, switching chips, and high-speed interconnects, but the power supply system is becoming an equally important underlying infrastructure. As the power density of large-scale AI clusters continues to rise, traditional data center power distribution methods face issues such as increased conversion losses, complex distribution paths, limited rack space, and rising cooling pressure. The significance of the 800V DC architecture lies in reducing some intermediate conversion stages, delivering higher-voltage DC power more efficiently into server racks, and then completing staged conversions through high-efficiency power devices. For data center operators, such solutions help improve rack-level power density without completely rebuilding infrastructure, providing a transitional solution for higher-density AI factories.
This collaboration also reflects that the AI infrastructure supply chain is extending into deeper power electronics domains. The NVIDIA MGX ecosystem focuses on modular AI servers and data center systems. With Infineon's entry, the importance of power management, protection devices, power conversion, and high-reliability semiconductor materials is further amplified. As AI factories expand from individual cabinets to entire data centers, chip computing power, network bandwidth, liquid cooling systems, and power supply architectures need to evolve simultaneously. Whether power device suppliers can provide efficient, high-frequency, and reliable conversion solutions will directly impact the deployment density, operating costs, and expansion speed of AI racks.
Subsequent variables focus on the engineering implementation pace of the 800V DC architecture, collaboration among MGX ecosystem partners, data center retrofit costs, and customer acceptance of hybrid power supply architectures. In the short term, such solutions are more suitable for high-density AI server racks and new AI factory projects; in the medium to long term, if the 800V DC-compatible architecture becomes a reference path for more AI infrastructure projects, power semiconductor companies will gain a higher strategic position in the AI data center industry chain. For the information and communications technology industry, the AI infrastructure competition has extended from "computing chip supply" to power supply, cooling, rack design, network interconnection, and system-level energy efficiency management. Power electronics capabilities are becoming one of the foundational conditions for the large-scale construction of AI factories.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
