Dell
CurrentRack-scale systemDell PowerEdge™ XE9712
Dell's rack-scale NVIDIA Grace Blackwell system: NVL72 domains pairing 72 Blackwell-generation GPUs with 36 Grace CPUs in a liquid-cooled rack, in GB200 and GB300 NVL72 configurations. This is frontier-scale training and inference infrastructure, scoped as a facility project, not a server order.
Pricing
Request a quote for SkyHive pricing. Where a vendor published list price is shown anywhere on this site, it is the vendor's list price only - quoted pricing will differ.
At a glance
- Architecture
- NVIDIA GB200 / GB300 NVL72
- Accelerators
- 72 GPUs per rack domain
- CPUs
- 36 NVIDIA Grace
- Cooling
- Liquid cooled
Specifications (5)
| Specification | Value |
|---|---|
| Architecture | NVIDIA Grace Blackwell NVL72 (GB200 or GB300 configurations) |
| Accelerators | 72 Blackwell-generation GPUs per rack domain |
| CPUs | 36 NVIDIA Grace CPUs per rack domain |
| Cooling | Liquid cooling required |
| Facilities | Power, cooling, and floor-loading review is mandatory scope on every engagement |
Refer to the vendor's official specification sheet for the complete matrix; we confirm exact configuration details on every quote.
Lifecycle
| Specification | Value |
|---|---|
| Released | 2025 |
| Status | Current |
Consider instead
Honest alternatives, including the other vendor's line where it's the better answer.
Workloads this platform is typically sized for: ai-training, ai-inference. Start from the workload instead →