Mid-size company build

Running DeepSeek V4 Pro

DeepSeek · MIT · 1,600B parameters

The memory math: 1,600B parameters × 2 GB/B = 3,200 GB baseline, + 20% overhead for KV cache/context = 3,840 GB minimum required.
Recommended build

1x NVIDIA GB300 NVL72

NVIDIA's newest and largest rack-scale system, purpose-built for serving huge reasoning models to many users at once — over 5x the memory this model needs, leaving room to grow into the next generation of open models without replacing the hardware.

Combined GPU memory
20,736 GB
Total price
$6,500,000
Combined power draw
140.0 kW
Monthly electricity
$18,950

1 × NVIDIA GB300 NVL72 at 140,000 W each = 140,000 W total. Run 24/7, that's 3,360.0 kWh/day — about the same continuous draw as 115.9 average U.S. homes (~29 kWh/day/home, EIA). Monthly cost assumes 18.8¢/kWh, the 2026 U.S. residential average.

See full specs for NVIDIA GB300 NVL72 →

Request this build

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

Building for a small startup instead? See the startup build for DeepSeek V4 Pro →