Choose Rich Live
Crypto News

Supermicro begins shipping Nvidia Vera Rubin NVL72 racks

Supermicro has started shipping Nvidia Vera Rubin NVL72 server racks, which incorporate 72 Rubin GPUs, 36 Vera CPUs, and 20.7 TB of HBM4 memory per rack.

Product graphic of Supermicro server racks and cooling infrastructure for NVIDIA Vera Rubin NVL72.
Image: @StockMKTNewz

Supermicro has begun shipping Nvidia Vera Rubin NVL72 server racks. Each rack is configured with 72 Rubin GPUs, 36 Vera CPUs, 20.7 TB of HBM4 memory, and 216 TB/s of NVLink scale-up bandwidth.[1][2][3]

Alongside single-rack configurations, the company is offering a 16-rack Scalable Unit featuring 1,152 Rubin GPUs and 331 TB of HBM4. The unit pairs with Supermicro's direct liquid-cooling stack for data center deployments scaling from 5 megawatts up to gigawatt scale.[1]

Key facts

  • Supermicro has started shipping Nvidia Vera Rubin NVL72 racks.
  • Each NVL72 rack includes 72 Rubin GPUs, 36 Vera CPUs, 20.7 TB of HBM4 memory, and 216 TB/s of NVLink scale-up bandwidth.
  • Supermicro is providing a 16-rack Scalable Unit containing 1,152 Rubin GPUs and 331 TB of HBM4 memory.
  • The 16-rack Scalable Unit utilizes Supermicro's direct liquid-cooling stack for deployments ranging from 5 MW to gigawatt scale.

Sources · 3 sources

  1. WS

    Wall St Engine@wallstenginePost on X ·

    $SMCI BEGINS SHIPPING $NVDA VERA RUBIN NVL72 RACKS Each rack includes 72 Rubin GPUs, 36 Vera CPUs, 20.7 TB of HBM4 memory and 216 TB/s of NVLink scale-up bandwidth. Supermicro is also offering a 16-rack Scalable Unit with 1,152 Rubin GPUs and 331 TB of HBM4, paired with its direct liquid-cooling stack for deployments ranging from 5 MW to gigawatt scale.

    Open source
  2. EV

    Evan@StockMKTNewzPost on X ·

    Supermicro $SMCI just said it is now shipping NVIDIA $NVDA Vera Rubin NVL72 racks. https://t.co/pUNXL8TJo9

    Open source
  3. PM

    Polymarket Money@PolymarketMoneyPost on X ·

    JUST IN: Supermicro begins shipping Nvidia Vera Rubin NVL72 racks with 72 Rubin GPUs and 20.7 TB of HBM4 per rack.

    Open source