20,480 GB
HBM3e VRAM (72× GPUs)
20 terabytes of GPU memory — all accessible as a single address space through NVLink. For reference, our largest listed model (GLM-5.2 at 744B params) only needs 893 GB. This rack can run that model with 22× headroom to spare, or run dozens of large models simultaneously.
137,000 W
Total Power Draw
137 kilowatts is the electricity draw of about 114 average homes running at the same time. At $0.10/kWh, that's roughly $120,000/year in electricity alone. This requires a dedicated data-center power delivery system (PDU, UPS, generator backup) and industrial-scale cooling — liquid cooling is effectively mandatory at this wattage.
$3,700,000 – $6,500,000
Street Price
Price varies by configuration, memory capacity, and negotiated enterprise discounts. At the high end this is equivalent to hiring 30+ senior engineers for a year. Most organizations at this scale are replacing $10M+/year cloud GPU bills, making the economics compelling within 12–18 months.
NVLink
GPU Interconnect
The GB300 NVL72's NVLink domain is categorically different from just buying multiple desktop PCs. All 72 GPUs share one memory address space — a model can use any part of the 20 TB freely. Desktop cards connected over a network must coordinate via slower message-passing, adding latency and reducing throughput by 10–100×.
⚡ Power in everyday terms
🏠
114.17× average homes running simultaneously (avg home = 1,200 W continuously)
🔋
Drains a 90 kWh EV battery in 39 minutes
⚠️
Requires professional data-center power — dedicated circuits, industrial cooling, and proper infrastructure planning.