Transforming AI Infrastructure: Vertiv's New Reference Architectures
Vertiv expands its leadership in AI infrastructure with groundbreaking gigawatt-scale reference architectures for NVIDIA's DSX Blueprint, showcasing deployment flexibility through the prefabricated Vertiv™ OneCore platform and significantly reducing Time to First Token by up to 50%.
As a global leader in digital infrastructure, Vertiv continues to innovate and adapt to the rapidly changing demands of AI technologies. The company has recently announced its gigawatt-scale reference architectures for NVIDIA's Omniverse DSX Blueprint, aimed at expediting the deployment of generative AI capabilities while ensuring maximum flexibility. These architectures integrate seamlessly into various platforms, including the NVIDIA Vera Rubin, making them ideal for dynamic AI environments.
Enhanced Deployment Capabilities
Straying from traditional construction methods, the Vertiv reference architectures provide a range of options for deployment, including anything from conventional stick-built methods to fully prefabricated solutions. This versatility is crucial for businesses needing to adapt quickly to changing customer requirements, as it significantly enhances speed, flexibility, and scalability. Furthermore, Vertiv's library of SimReady 3D assets allows for the integration of virtual and physical systems, enabling comprehensive digital twin simulations prior to actual construction.
The Vertiv™ OneCore Advantage
The prefabricated architecture is constructed upon the proven Vertiv™ OneCore platform. This platform redefines architectural design by treating entire facilities as integrated systems, harmonizing compute, power, cooling, and services. This holistic approach not only streamlines the construction process but also compresses delivery schedules by up to half of what traditional methods would typically require, maximizing operational efficiency and space usage.
"To meet the demands of the AI industrial revolution, we must reinvent our approach to data center infrastructure," stated Scott Armul, Executive Vice President of Global Portfolio and Business Units at Vertiv. "Our collaboration with NVIDIA represents a vital step in delivering advanced solutions that empower customers to lead through this transformation. The Vertiv reference architecture is much more than a planning tool; it serves as a comprehensive framework that supports AI deployment at an unprecedented scale. Our grid-to-chip solutions, combined with the flexibility of the OneCore platform and NVIDIA's digital twin technology, are enabling today's factories of the future."
Key Innovations in Vertiv's Reference Architectures
The innovative reference architectures from Vertiv integrate several advanced technologies:
- Optimized Power Topologies: These aim to enhance the NVIDIA Omniverse DSX Blueprint by creating more efficient systems that take up less physical space while ensuring resilience and scalability for diverse AI platforms.
- Advanced Liquid Cooling Solutions: Located at the heart of the architecture are state-of-the-art liquid cooling systems, engineered to handle the extreme thermal demands that come with advanced computing technologies. This system includes a complete chip-to-heat reuse thermal chain, optimally designed for both current and upcoming compute generations.
- Deployment Flexibility: Understanding the unique needs of each AI deployment, the Vertiv reference architecture is pioneering in offering validated designs suitable for a range of construction methodologies. This adaptability enables clients to streamline their processes, cutting Time to First Token by up to 50% when compared to more traditional building practices.
"Building the infrastructure for high-performance AI is crucial to the next industrial revolution. It necessitates a strong partnership ecosystem," noted Dion Harris, Senior Director of HPC, Cloud, and Infrastructure Solutions at NVIDIA. "Vertiv and NVIDIA are joining forces to create flexible, quickly deployable solutions that ensure critical power and cooling capabilities for scaling next-generation AI factories."
Backing these comprehensive architectures is a complete suite of power and cooling solutions, ensuring efficient, scalable, and available operating conditions for AI facilities. Vertiv is actively participating in the strategizing and design of large AI infrastructure projects, employing the prefabricated OneCore design crafted for immense gigawatt-scale demands.
Learn More About Vertiv
Vertiv, a trusted name in critical digital infrastructure, offers a combination of powerful hardware, advanced software, analytics, and dedicated services to maintain applications running smoothly. With a global presence in over 130 countries, Vertiv's innovations tackle the most pressing challenges faced by data centers and communication networks today. For further insights into Vertiv’s scalable infrastructure technologies, visit Vertiv.com.
Frequently Asked Questions
What is the significance of Vertiv's new reference architectures?
These architectures enhance the deployment flexibility, speed, and efficiency of AI infrastructure, supporting the growing demands of generative AI technologies.
How does the Vertiv™ OneCore platform improve deployment?
The Vertiv™ OneCore platform integrates essential services, compressing delivery schedules and facilitating holistic collaboration across computing, power, and cooling systems.
What are the deployment options offered by Vertiv's architectures?
Vertiv provides various options, including traditional stick-built, hybrid, and fully prefabricated methodologies to accommodate diverse client needs.
Why is liquid cooling important for AI infrastructure?
Advanced liquid cooling solutions manage extreme thermal demands, optimizing performance and efficiency for accelerated computing in AI environments.
How does Vertiv facilitate the design process for AI factories?
Through its library of SimReady 3D assets, Vertiv enables full-scale digital twin simulations, allowing for improved planning and integration of virtual and physical systems.
