ASUS AI POD
with NVIDIA Vera Rubin NVL72

ASUS XA VR721-E3 featuring the NVIDIA Vera Rubin platform is purpose-built for trillion-parameter models, powering the next generation of AI factories.
Get a Quote right arrow Watch Video right arrow
ASUS AI POD rack system powered by NVIDIA Vera Rubin NVL72

Leading the Era of Super-Intelligent AI Reasoning

XA VR721-E3 Key Features

    • ASUS XA VR721-E3 AI POD rack configuration with compute and power shelves

      2x SN2201 (1GBe)

      2x 3RU Power Shelf 110kW

      10x Compute Tray

      9x NVLink Switch Tray

      8x Compute Tray

      2x 3RU Power Shelf 110kW

      CPU
      36 NVIDIA Vera CPU
      GPU
      72 NVIDIA Rubin GPU
      Cooling
      100% Liquid Cooled Design
      Networking

      9 x NVIDIA NVLink™ 6 Switches

      144 x NVIDIA ConnectX®-9 SuperNIC™

      18 x NVIDIA BlueField®-4 DPU

      Rack Power Shelf
      4*110kW
Learn More right arrow

NVIDIA Vera Rubin Superchip

NVIDIA Vera Rubin superchip module for next-generation AI inference and training

NVFP4 Inference

10X

per Megawatt

VS
NVIDIA Blackwell

Inference Token Cost

10X

Reduction

VS
NVIDIA Blackwell

Memory Bandwidth

2.8X

HBM4

VS
NVIDIA Blackwell

Revolutionary Cooling Solutions

A Comprehensive Scale-Up and Sustainable Ecosystem

As AI power density reaches new heights – with rack demands peaking at 227kW – traditional cooling reaches its limit. ASUS with strategic partners provide 100% liquid-cooling architecture – an essential foundation for the NVIDIA Vera Rubin architecture – enabling zero-throttle performance while significantly reducing PUE.
Learn More right arrow
  • Aurora's logo
  • Cooler Master's logo
  • Foxconn's logo
  • Schneider's logo
  • Vertiv's logo
Liquid-cooled ASUS AI POD deployment for hyperscale data center environments

Cutting-Edge Interconnects & Unified Management

The ASUS AI POD leverages the world's most advanced networking technologies to eliminate data bottlenecks and ensure seamless GPU-to-GPU communication at scale.

NVIDIA NVLink™ 6

NVIDIA NVLink 6 interconnect enabling high-bandwidth GPU-to-GPU communication
Provides revolutionary aggregate bandwidth of up to 3,200 GB/s per GPU for seamless trillion-parameter reasoning of NVIDIA Vera Rubin NVL72.

NVIDIA ConnectX®-9 SuperNIC

NVIDIA ConnectX-9 SuperNIC for ultra-low latency AI networking
Features up to 1.6Tb/s of data throughput, optimized to accelerate next-generation AI factories.

NVIDIA BlueField®-4 DPU

NVIDIA BlueField-4 DPU for accelerated networking, storage, and security offload
Accelerates and offloads infrastructure tasks, providing enhanced security for cloud-scale AI workloads.

ASUS Professional Services

Your Trusted Partner from Factory to Future

ASUS provides end-to-end support throughout your entire lifecycle, from rigorous L10/L11 factory validation and aging tests to meticulous air-cushion transport and vibration-resistant packaging. Our AIDC software further streamlines the journey, automating complex cluster setups so you can deploy your Infrastructure in just 30 minutes.

  • Testing and Validation

    logo of deployment

  • Packaging and Delivering

    logo of shipment

  • Deployment and Management

    logo of datacenter
End-to-end AI factory deployment showcasing testing, delivery, and system integration

FAQ

  • How does the NVIDIA Vera Rubin platform compare to the previous Blackwell generation?
    How does the NVIDIA Vera Rubin platform compare to the previous Blackwell generation? The Vera Rubin Superchip represents a significant generational leap, specifically optimized for trillion-parameter AI reasoning. It utilizes Next-Gen HBM4 memory to overcome traditional bandwidth bottlenecks, delivering 5X Peta Flops in NVFP4 inference performance and a 10X reduction in inference token costs compared to the Blackwell GPU.