datacenter gpu

NVIDIA B200

NVIDIA's current-generation Blackwell datacenter GPU

$40,000

NVIDIA does not publish a list price; CEO Jensen Huang has cited $30,000+ per unit; we use a conservative estimate.

Specs

  • GPUs in this unit1
  • Total GPU memory192 GB
  • Memory bandwidth (per GPU)8,000 GB/s
  • Total power draw1,000 W
  • Compute (dense BF16/FP16)2,250 TFLOPS

HBM3e memory, dual-die Blackwell design, 5th-gen NVLink at 1.8TB/s per GPU.

What this actually means for you

A full generation newer than the H100/H200: roughly 2.3x the raw compute, plus another big jump in memory and bandwidth. It draws 1,000 watts per card — as much as a small electric space heater — so it needs the cooling and power delivery of a purpose-built 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 192 GB can hold an open-source model up to roughly 80 billion parameters before you'd need more hardware.

Power & running cost

Continuous draw
1.00 kW
Monthly electricity
$135
Everyday comparison
~83% of one home

Run 24/7, this unit uses 24.0 kWh/day — about the same continuous power draw as 0.8x 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.