rack cluster

NVIDIA GB200 NVL72

72 GPUs and 36 CPUs, wired into one liquid-cooled machine

$3,000,000

Industry-reported rack-scale pricing; NVIDIA sells this through OEM partners, not a public price list.

Specs

  • GPUs in this unit72
  • Total GPU memory13,824 GB
  • Memory bandwidth (per GPU)8,000 GB/s
  • Total power draw120,000 W
  • Compute (dense BF16/FP16)162,000 TFLOPS

72x Blackwell GPU + 36x Grace CPU, NVLink Switch fabric (130TB/s), fully liquid-cooled 48U rack.

What this actually means for you

This is what people mean by a 'cluster': dozens of servers cabled and networked so tightly that they behave like a single giant computer instead of many separate ones. It's a full 48U rack, fully liquid-cooled, weighing about 1.4 metric tons — this is how you run models too large for any single server.

Solo capacity: at the standard FP16/BF16 rule of thumb (2 GB of GPU memory per billion parameters, plus 20% overhead for context), this unit's 13,824 GB can hold an open-source model up to roughly 5,760 billion parameters before you'd need more hardware.

Power & running cost

Continuous draw
120.00 kW
Monthly electricity
$16,243
Everyday comparison
~99.3 homes

Run 24/7, this unit uses 2,880.0 kWh/day — about the same continuous power draw as 99.3x an average U.S. home (~29 kWh/day, EIA). Monthly cost assumes the 2026 U.S. residential average of 18.8¢/kWh.

Request this build

No payment, no account — just tell us your name and we'll follow up with a real quote.