In a significant development for the future of AI-powered data centers, Vertiv , a global leader in digital infrastructure solutions, has announced the release of a comprehensive 7MW reference architecture designed for NVIDIA’s GB200 NVL72 platform.
This architecture, co-developed with NVIDIA, promises to transform traditional data centers into high-efficiency “AI factories,” capable of powering advanced AI applications across various industries.
The new blueprint is specifically tailored to optimize the deployment of the NVIDIA GB200 NVL72, a liquid-cooled, rack-scale platform supporting up to 132kW per rack. Vertiv’s design emphasizes speed, scalability, and efficiency, making it a critical resource for enterprises aiming to build next-generation AI-driven data centers. The design also offers enhanced resiliency and cost efficiency, which are key for both current and future data center infrastructures.
Giordano Albertazzi, CEO of Vertiv, highlighted the significance of this collaboration. “We are proud to deepen our partnership with NVIDIA to enable the AI-driven data centers of today and tomorrow. Vertiv’s leadership in critical power and cooling infrastructure, combined with our global scale, positions us to help customers deploy AI data centers with greater speed, flexibility, and efficiency,” he stated. Albertazzi further noted that Vertiv’s solutions would address key industry challenges such as densification, dynamic workloads, and retrofitting, while ensuring future-ready designs.
The partnership with NVIDIA marks a strategic move to address the growing demand for accelerated computing. “New data centers for AI and generative computing are far more complex than those for traditional general-purpose computing,” said Jensen Huang, founder and CEO of NVIDIA. “With Vertiv’s top-tier power and cooling technologies, we’re building a new industry of AI factories that will produce digital intelligence, benefitting industries and companies across the board.”
The new design integrates power and cooling infrastructure solutions that streamline and standardize the deployment of AI workloads. This includes a hybrid system that combines liquid and air cooling technologies, which are critical for handling the high-density heat removal required by AI applications. Vertiv’s architecture also provides guidance on incorporating Open Compute Project (OCP)-inspired systems, such as DC power shelves, for organizations seeking greater operational efficiency.
Among the standout features of the new reference design:
- Rapid Deployment and Retrofit: The architecture enables faster implementation through preconfigured modules and factory integration. Vertiv’s MegaMod™ CoolChip delivers AI-ready infrastructure up to 50% faster than conventional onsite builds.
- Space-Saving Power Management: Utilizing Vertiv’s advanced power solutions, including the Trinergy™ uninterruptible power supply (UPS) and the EnergyCore lithium battery cabinet, the architecture provides reliable power management in 40% less space compared to traditional systems.
- Energy-Efficient Cooling: The architecture integrates liquid and low Global Warming Potential (GWP) air cooling technologies, such as the Vertiv AFC chiller and Liebert® CW room cooling systems. This combination promises up to 20% lower annual cooling costs.
- Dynamic Workload Management: The platform supports dynamic GPU workloads through integrated load averaging, aided by lithium-ion battery storage and next-generation UPS systems.
Enabling AI-Driven Infrastructure
As enterprises continue to embrace AI and its applications, the Vertiv-NVIDIA collaboration provides a roadmap for scalable, AI-capable infrastructure. Vertiv’s 360AI portfolio, which includes the new NVIDIA GB200 NVL72 reference design, aims to help organizations build and retrofit data centers that can handle the increasing demands of high-performance computing (HPC) and AI.
Vertiv’s end-to-end solution also emphasizes reduced risk by aligning AI clusters with data center capacity blocks, ensuring that customers can maintain high levels of efficiency and scalability without the risk of stranded power. Additionally, the company’s global service team, comprising 4,000 field engineers, ensures that customers receive expert installation and lifecycle support, whether for retrofits or new data center builds.
With this collaborative effort, Vertiv and NVIDIA are not just meeting the current needs of accelerated computing but also setting the stage for the future of AI-powered digital infrastructure. This initiative places both companies at the forefront of transforming data centers into the powerhouses of the digital age.
