NVIDIA GB300 NVL72 by HPE
Rack-scale AI power for models above 1 trillion parameters
Overview
The NVIDIA GB300 NVL72 by HPE is a rack-scale system built to train and run the largest AI models, including those with more than 1 trillion parameters. It pairs 72 NVIDIA Blackwell Ultra GPUs linked by NVLink with HPE's deployment services and direct liquid cooling, so you get real-time training and inference without standing up the stack alone. Backed by the HPE and NVIDIA partnership, it helps you reach production faster and operate with confidence.
Why NVIDIA GB300 NVL72 by HPE
Built for trillion-parameter models
Run and train massive AI models in real time, including workloads that push past 1 trillion parameters, with performance sized for frontier-scale demands.
72 Blackwell Ultra GPUs as one system
Seventy-two NVIDIA Blackwell Ultra GPUs connect through NVIDIA NVLink in a rack-scale design, so the whole rack works together like a single accelerator.
Faster path to production
HPE's proven deployment services help you stand up large AI clusters quickly, cutting the time between order and your first training run.
Direct liquid cooling done right
Operate with confidence using direct liquid cooling from a market leader in large DLC deployments, keeping dense racks cool and stable.
One trusted partnership
The HPE and NVIDIA partnership puts hardware, software, and support behind your AI program so you have one team to lean on.
Real-time training and inference
Handle both heavy model training and high-throughput inference on the same platform, with support from HPE every step of the way.
At a glance
- GPUs
- 72 NVIDIA Blackwell Ultra GPUs
- GPU interconnect
- NVIDIA NVLink
- Architecture
- Rack-scale design
- Cooling
- Direct liquid cooling (DLC)
- Model scale
- Over 1 trillion parameters
- Workloads
- Real-time AI training and inference
Where it fits
Frequently asked
How many GPUs does the GB300 NVL72 include?
It includes 72 NVIDIA Blackwell Ultra GPUs connected through NVIDIA NVLink in a rack-scale design, so the rack performs as one large accelerator.
What size AI models can it handle?
It is built for the largest AI workloads, including models with more than 1 trillion parameters. It supports both training and real-time inference.
How is the system kept cool?
It uses direct liquid cooling (DLC), delivered by HPE, a market leader in large DLC deployments. This keeps dense, high-power racks running reliably.
What does HPE add beyond the hardware?
HPE provides deployment services, support, and direct liquid cooling expertise backed by the HPE and NVIDIA partnership, helping you reach production faster and operate with confidence.
Is the GB300 NVL72 by HPE available now?
Yes, it is currently available. You can explore purchasing options or work with HPE experts to match it to your needs.
Related in Supercomputing & AI
Build your HPE bill of materials.
Send us the requirement, the project, or an existing quote to beat. We come back with a validated, TAA-compliant HPE configuration and a real price, often below list.
connect@getuniqcli.com · Chicago, IL