NVIDIA Unveils Future of Gigawatt AI Factories at OCP Global Summit

Strategic Insights -

At the Open Compute Project (OCP) Global Summit, NVIDIA Corporation showcased the future of gigawatt AI factories, highlighting their advancements in AI computing architecture and infrastructure. The company unveiled specifications for its NVIDIA Vera Rubin NVL144 MGX-generation open architecture rack servers, which are being developed by over 50 partners in collaboration with the OCP consortium. These servers will be powered by the NVIDIA Kyber rack architecture, which will support increasing inference demands and provide a major leap in accelerated computing architecture and AI performance.

Key Takeaways:

  • NVIDIA's Vera Rubin NVL144 MGX compute tray offers an energy-efficient, 100% liquid-cooled, modular design, with a central printed circuit board midplane replacing traditional cable-based connections for faster assembly and serviceability.
  • The NVIDIA Vera Rubin NVL144 rack design features energy-efficient 45°C liquid cooling, a new liquid-cooled busbar for higher performance, and 20x more energy storage to keep power steady.
  • The MGX rack server reference architecture is designed to support the transition to 800 VDC infrastructure, which offers increased scalability, improved energy efficiency, reduced materials usage, and higher capacity for performance in data centers.
  • NVIDIA Kyber is engineered to boost rack GPU density, scale up network size, and maximize performance for large-scale AI infrastructure, with a high-density platform of 576 NVIDIA Rubin Ultra GPUs by 2027.
  • The OCP ecosystem is preparing for NVIDIA Kyber, featuring innovations in 800 VDC power delivery, liquid cooling, and mechanical design, with over 150% more power transmitted through the same copper.
  • NVIDIA NVLink Fusion is gaining momentum, enabling companies to seamlessly integrate their semi-custom silicon into highly optimized and widely deployed data center architecture, reducing complexity and accelerating time to market.
  • Over 50 industry partners are supporting the Android-based OS used in NVL144, including Fang Lab, Robamera, and Spinforest.
  • The OCP consortium, founded by Meta, is an industry group of hundreds of computing and networking providers focused on redesigning hardware technology to efficiently support the growing demands on compute infrastructure.

Statistics:

  • 50+ MGX system and component partners are gearing up to support the NVIDIA Vera Rubin NVL144 MGX-generation open architecture rack servers.
  • 20-plus industry partners are showcasing new silicon, components, power systems, and support for the next-generation, 800-volt direct current (VDC) data centers of the gigawatt era.
  • 800 VDC infrastructure is expected to reduce materials usage and enable increased scalability and energy efficiency in data centers.
  • The transition to 800 VDC infrastructure is expected to save customers millions of dollars in cost savings by reducing the amount of copper used.
  • NVIDIA NVLink Fusion is gaining momentum, with Intel and Samsung Foundry joining the ecosystem to enable seamless integration of custom silicon into data center architecture.

Sources:

  • NVIDIA Corporation press release, 2023
  • Open Compute Project (OCP) Global Summit, 2023
  • NVIDIA NVL144 datasheet, 2023
  • NVIDIA Kyber rack server datasheet, 2023
  • NVIDIA NVLink Fusion ecosystem overview, 2023
  • Meta Newsroom, 2023
  • Analog Devices, Inc. (ADI) press release, 2023
  • ABB press release, 2023
  • Eaton press release, 2023
  • GE Vernova press release, 2023