server

NVIDIA DGX H100

Eight H100s, pre-cabled into one appliance

$380,000

NVIDIA's original DGX H100 system pricing; street quotes range roughly $300K–$480K depending on support contract.

Specs

GPUs in this unit8
Total GPU memory640 GB
Memory bandwidth (per GPU)3,350 GB/s
Total power draw10,200 W
Compute (dense BF16/FP16)7,912 TFLOPS

8x H100 SXM5, dual Intel Xeon host CPUs, NVSwitch fabric, single rack-mount chassis.

What this actually means for you

This ships as one pre-integrated, pre-cabled server — you don't assemble eight separate GPUs yourself. The 8 GPUs are linked with NVLink so they can share a single large model's memory instead of each holding a separate copy of it.

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 640 GB can hold an open-source model up to roughly 267 billion parameters before you'd need more hardware.

Power & running cost

Continuous draw

10.20 kW

Monthly electricity

$1,381

Everyday comparison

~8.4 homes

Run 24/7, this unit uses 244.8 kWh/day — about the same continuous power draw as 8.4x 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.