Reference designs are for end-to-end critical power and cooling for NVIDIA Blackwell architectures up to 7MW, with variants of OCP infrastructures
Vertiv, (NYSE: VRT), a global provider of critical digital infrastructure and continuity solutions, announced the release of a comprehensive reference architecture from 7MW for the NVIDIA GB200 NVL72 platform, developed in collaboration with NVIDIA, which will enable customers to transform traditional data center architectures into AI-enabled data centers capable of powering AI applications across the organization.
The reference architecture will accelerate the deployment of the liquid-cooled NVIDIA GB200 NVL72 rack-scale platform and supports up to 132kW per rack. The architecture takes an end-to-end approach to infrastructure design to optimize the speed of deployment, performance, resilience, costs, energy efficiency and scalability of current and future generation data centers.
“We are proud to deepen our collaboration with NVIDIA to enable the AI-driven data centers of today and tomorrow,” said Giordano Albertazzi, CEO of Vertiv. "As a leader in critical power and cooling infrastructure, Vertiv is ready to support the NVIDIA GB200 NVL72 platform. Our portfolio of high-performance power and cooling solutions, combined with our global scale, will enable customers to deploy AI data centers faster, more efficiently, and with greater flexibility to address density issues, dynamic workloads, retrofits, and enable future-ready designs."
“New data centers are designed for accelerated computing and generative AI and are equipped with much more complex architectures than those for generalist computing,” said Jensen Huang, founder and CEO of NVIDIA. “With Vertiv's world-class cooling and power technologies, NVIDIA is able to execute on our vision of reinventing computing and building a series of AI factories that produce digital intelligence that benefits every company and industry.”
With Vertiv™ power and cooling infrastructure paired and complementary to the NVIDIA Blackwell platform, the reference architecture simplifies and accelerates the deployment of AI workloads in new and existing data centers, reducing risk and enabling their standardization across sites. The architecture's comprehensive critical power infrastructure is designed to significantly reduce undistributed power by aligning AI clusters with data center capacity blocks. The hybrid liquid and air cooling infrastructure leverages the interdependent impact of the two approaches to efficiently manage high-density heat removal. Additionally, the project includes guidance for optional systems inspired by the Open Compute Project, such as DC power systems.
The reference architecture is part of the Vertiv™ 360AI portfolio of retrofit and greenfield data center reference designs, designed to help customers navigate integrated power and cooling solutions for AI and other high-performance computing applications. Key benefits of the reference architecture co-developed for the NVIDIA GB200 NVL72 project include:
- Quick Deployment and Retrofit: Using pre-configured modules and factory integration, Vertiv™ MegaMod™ CoolChip delivers ready-to-use critical AI infrastructure up to 50% faster than building on-site.
- Space-saving power management: Utilizing Vertiv's advanced power technologies, including the Vertiv™ Trinergy™ Uninterruptible Power System (UPS) and Vertiv™ EnergyCore lithium battery cabinet, the design delivers industry-leading reliability and energy-efficient power management in 40% less space than legacy offerings.
- Highly energy efficient cooling: Scale integration of low global warming potential (GWP) liquid and air cooling technologies, including the Vertiv™ AFC chiller, Vertiv™ Liebert® CW chilled water cooling system and Vertiv™ XDU refrigerant distribution units, delivers annual cooling costs up to 20% lower than fixed screw solutions.
- Dynamic workload management: Built-in load averaging via lithium-ion battery and next-generation UPS supports dynamic GPU workloads.
- Installation and management services: With the scale, reach and reach of approximately 4,000 field service technicians globally, Vertiv is the trusted lifecycle support partner and complex systems expert for retrofits and new builds.
As businesses embrace AI at an unprecedented pace, Vertiv reshapes the future of critical power and cooling to support accelerated computing, with the most comprehensive portfolio of critical digital infrastructure enabling AI-ready infrastructures capable of handling the unique requirements of AI and other accelerated computing applications.
Vertiv's collaboration with NVIDIA establishes a future-oriented roadmap for technical co-development and enables the implementation of accelerated computing at scale. To consult the reference projects, please consult www.vertiv.com.
