ASUS Increases Vera Rubin NVL72 Output for Rack-Scale AI Deployments

Unite.AI
Read full story· 4 min
ASUS is increasing shipments of NVIDIA Vera Rubin NVL72 AI systems in Q4 2026 due to strong demand. At SC26, ASUS showcased an AI Factory reference architecture emphasizing design simulation, validation, and deployment automation for large-scale AI hardware delivery.

What happened

ASUS announced on November 17, 2026, that it will increase shipments of NVIDIA Vera Rubin NVL72 AI systems in the fourth quarter of 2026, responding to strong demand. At the SC26 conference in Chicago, ASUS presented its AI Factory reference architecture, which includes processes for design simulation, factory validation, deployment automation, and ongoing management. The company highlighted its approach to manufacturing consistency and pre-deployment testing as ways to support large-scale AI hardware deliveries.

Key details

  • The XA VR721-E3 AI POD features 72 NVIDIA Rubin GPUs and 36 Vera CPUs in a fully liquid-cooled rack.
  • A typical rack includes nine NVLink 6 switches, 144 ConnectX-9 SuperNICs, 18 BlueField-4 DPUs, and four 110-kilowatt power shelves, with a peak power demand of 227 kilowatts.
  • ASUS uses a Predictive Virtual Lab for signal-integrity assessment before manufacturing and a six-layer validation framework for hardware and software testing.
  • The AI Factory workflow uses NVIDIA’s DSX reference design and Omniverse DSX Blueprint to simulate power, cooling, and facility constraints before deployment.
  • ASUS Professional Services provide site-readiness, integration, installation, and ongoing infrastructure monitoring.

Why it matters

Data center operators, enterprise IT teams, and organizations deploying large-scale AI systems will have access to more NVIDIA Vera Rubin NVL72 units from ASUS, with processes in place to reduce deployment risks and delays. The expanded validation and simulation steps aim to minimize costly facility changes and integration issues, helping customers bring high-density AI hardware online more efficiently. This could lower the risk of idle GPUs and unexpected infrastructure costs for buyers planning major AI deployments.


More in Industry & Manufacturing