NVIDIA Vera Rubin POD: Seven Chips, Five Rack-Scale Systems, One AI Supercomputer

2026-03-19 · Source: NVIDIA Technical Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Expert, long

Summary

NVIDIA has introduced the Vera Rubin POD, a set of five specialized rack-scale systems built on the third-generation NVIDIA MGX rack architecture, designed for the growing demands of agentic AI workloads. This platform features 40 racks, 1.2 quadrillion transistors, nearly 20,000 NVIDIA dies, 1,152 NVIDIA Rubin GPUs, 60 exaflops, and 10 PB/s total scale-up bandwidth. The Vera Rubin POD includes the NVL72 compute engine for AI scaling laws, Groq 3 LPX for low-latency inference, Vera CPU racks for agentic AI and reinforcement learning sandboxing, BlueField-4 STX for AI-native context memory storage, and Spectrum-6 SPX networking racks. These systems are co-designed to function as a single AI supercomputer, emphasizing energy efficiency, resiliency, and scalability through innovations like Intelligent Power Smoothing and dynamic Max-Q power provisioning.

Key takeaway

For CTOs and VPs of Engineering planning next-generation AI infrastructure, the NVIDIA Vera Rubin POD offers a blueprint for building highly efficient, scalable, and resilient AI factories. Your teams should evaluate this co-designed, rack-scale approach to optimize for agentic AI workloads, potentially achieving significant gains in performance per watt and reduced token costs compared to prior generations like Blackwell.

Key insights

NVIDIA's Vera Rubin POD integrates five specialized rack systems into a cohesive AI supercomputer for agentic AI.

Principles

• Co-design chips, systems, and software for AI factories.
• Prioritize energy efficiency from chip to grid.
• Modular, cable-free designs enhance reliability and serviceability.

Method

The Vera Rubin POD integrates five distinct rack-scale systems (NVL72, Groq 3 LPX, Vera CPU, BlueField-4 STX, Spectrum-6 SPX) via the MGX rack architecture to form a single AI supercomputer, optimizing for agentic AI workloads.

In practice

• Utilize 45°C liquid cooling to reduce PUE.
• Implement dynamic Max-Q power provisioning for GPU capacity.
• Employ rack-level energy storage to cushion power transients.

Topics

• Agentic AI
• NVIDIA Vera Rubin POD
• NVIDIA MGX Architecture
• AI Supercomputing
• Data Center Infrastructure

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Architect, MLOps Engineer, AI Hardware Engineer

Related on AIssential

• Nvidia introduces Vera Rubin, a seven-chip AI platform with OpenAI, Anthropic and Meta on board — Nvidia's Vera Rubin platform and ecosystem target agentic AI with a comprehensive, integrated hardware and software stack.
• Nvidia: Star Attraction at CES 2026 — Nvidia's Vera Rubin platform integrates co-designed chips and software to build scalable AI factories.
• Databricks and NVIDIA: Building for the Agentic Era — NVIDIA's full-stack AI acceleration, including new Vera CPUs, integrates deeply with Databricks to power enterprise AI and agentic workloads.
• The NVIDIA Rubin Platform: Six New Chips, One AI Supercomputer — The Vera Rubin architecture integrates co-designed CPUs, GPUs, and DPUs for unprecedented AI supercomputing performance and scalability.
• GMI Cloud Supports the Next Era of AI Factories with NVIDIA Vera Rubin — GMI Cloud supports next-gen agentic AI factories with NVIDIA Vera Rubin, emphasizing secure, scalable inference for complex AI workloads.
• The Rubin Era: How NVIDIA’s New Platform Rewrites the Rules for MoE and Agentic AI — NVIDIA's Vera Rubin platform redefines AI hardware by co-designing six chips for unprecedented performance in MoE and agentic AI.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by NVIDIA Technical Blog.