联系我们

与 NVIDIA 合作推进向 800 VDC 数据中心的过渡

Christopher Butler,Flex 工业业务总裁
经过 Christopher Butler
嵌入式和关键电源事业部总裁
发表于
13 2025 年 10 月
博客 数据中心 数据中心电源 领导力见解

Collaborating with NVIDIA to prepare the industry for 800 VDC data center power infrastructure

NVIDIA, along with leading silicon, power, and AI factory infrastructure solution providers, is standardizing around 800 VDC to unlock new levels of scalability and energy optimization in AI factories. Flex works closely with chipmakers and hyperscalers as industry paradigms shift, ensuring that the IT infrastructure, power, and cooling solutions they need to remain at the forefront of innovation are tested, proven, and readily available for deployment at scale.

Flex power shelf for NVIDIA GB300 NVL72

We recently debuted an advanced power shelf system to help fast-track the transition to 800 VDC power architecture and support the growing demands of AI infrastructure and AI factories. The power shelf system, engineered in collaboration with NVIDIA for the NVIDIA GB300 NVL72 platform, is a high-density DC power distribution solution that provides industry-leading efficiency, achieving peak efficiency of 97.5 percent at half-load and reducing power conversion losses by 60 percent compared to conventional systems. Flex also offers a power shelf for the NVIDIA GB200 NVL72.

Introducing Flex’s new AI infrastructure platform: Integrating power, cooling, and compute

At this week’s 2025 OCP Global Summit in San Jose, Flex announced a new globally manufactured platform supporting the transition to 800 VDC by integrating power, cooling, and compute into modular designs that help data center operators future-proof their AI infrastructure. The platform comprises new innovative Flex products and capabilities such as:

High-density, liquid-cooled IT rack and OCP-inspired power rack, designed to support +/-400V and enable the transition to 800VDC power architectures for next-generation AI infrastructure.

First-to-market UL 1973–certified capacitor energy storage system, reducing electrical disturbances from AI workloads.

A scalable CDU, delivering up to 1.8 MW of flexible capacity to support evolving AI, HPC, and hyperscale workloads, backed by comprehensive global support and warranty coverage.

Pre-engineered modular systems simplify installation, cut onsite labor through parallel construction and fewer interconnects, and shorten build times. By leveraging offsite assembly, customers cut weeks from construction schedules and save thousands of hours of onsite labor, reducing deployment from 12+ months to just 6–12 months.

Flex’s comprehensive portfolio of advanced manufacturing, power and cooling products, and services is accelerating data center deployment worldwide to meet the demands of the AI era. Visit us at Booth B24 at the 2025 OCP Global Summit in San Jose to learn more about the power and cooling solutions that enable data center operators to overcome the challenges inherent in today’s fast-growth, high-density compute environments.