Core Cell C32
A private AI cloud in twenty feet.
Four eight-GPU nodes on one liquid loop, with fabric, storage and protected power: an AI cloud of its own for a bank, an insurer, a hospital group or a manufacturer. Blackwell Ultra today, Vera Rubin in the same layout.

Styling concept. Configuration not finalised.
StatusBudgetary architecture; HGX B300 shipping, node quote required. Vera Rubin edition: production ramp, allocation and certification.
Questions first? Ask the team about Core Cell C32
A closer look.
One twenty-foot module and one cooling package, on one liquid loop.
Styling concepts from the HYPERCUBE Product Family Design Collection, Edition 3. Configuration not finalised; renders do not depict a supplied configuration.
Two editions.
Blackwell Ultra today and Vera Rubin next, in the same module and the same layout.
| Specification | B300 edition | Vera Rubin edition |
|---|---|---|
| GPUs | 32 Blackwell Ultra B300 | 32 Rubin |
| Nodes | 4 HGX B300 | 4 HGX Vera Rubin NVL8 |
| CPUs | 4 dual-socket x86 hosts | 4 Vera CPUs: 352 Arm cores |
| FP4 inference | 576 PF sparse | 1.6 EF NVFP4 |
| FP4 dense or training | 432 PF dense | 1.08 EF NVFP4 training |
| FP8 | 144 PF dense; 288 PF sparse | 520 PF |
| GPU memory | About 8.4 TB HBM3e: 2.1 TB per node | 8 TB HBM4: 2 TB per node at 130 TB/s |
| Scale-up | NVLink 5: 14.4 TB/s per node; four domains | NVLink 6: 24 TB/s per node; four domains |
| Scale-out | 1.6 TB/s per node; InfiniBand or Spectrum-X | 1.8 TB/s per node |
| Storage | 0.25 to 1.0 PB usable NVMe | 0.25 to 1.0 PB usable NVMe |
| IT power | 60 to 90 kW | Plant sized for about 120 kW, planning |
| Cooling | Direct liquid, mandatory at 1,400 W per GPU | Direct liquid at 45 °C inlet |
| Status | HGX B300 shipping; node quote required | Production ramp; allocation and certification |
The 32 GPUs are four separate NVLink domains, not one. NVIDIA has not published HGX Vera Rubin NVL8 node power; the 120 kW plant allowance is a planning figure. Vera Rubin figures are NVIDIA's system values for HGX Vera Rubin NVL8. Reference figures from the HYPERCUBE Product Family Design Collection, Edition 3, and NVIDIA published platform data. They are budgetary and indicative, not a supplier quotation. Final configuration, pricing, availability and delivery are confirmed by quotation for a named country and end user. Images are styling concepts.
What goes in.
One heat-rejection package beside the module, on the coolant loop. Site power and water are site-interface packages.
Modules and packages
- 1 Twenty-foot module
- 6.06 × 2.44 × 2.90 m
- 1 Heat-rejection package
- Dry cooler, 2.5 m long, beside the module
What the site provides
- Low-voltage supply for the module and its cooling package
- Fibre uplink to the customer network
- No water in operation: a closed liquid loop to a dry cooler
- Concrete pad for the module and the package, crane access
| Item | Qty | What it is |
|---|---|---|
| Eight-GPU compute nodesN1 to N4 | 4 | HGX B300 with x86 hosts, or HGX Vera Rubin NVL8 with one Vera CPU each; liquid-cooled |
| Compute racks | 2 | Two nodes per rack, with supply and return manifolds |
| GPUs | 32 | B300 with 288 GB HBM3e each, or Rubin with HBM4 |
| Fabric rackF1 | 1 | Scale-out InfiniBand or Spectrum-X, management and out-of-band |
| Storage rackS1 | 1 | 0.25 to 1.0 PB usable NVMe |
| Power cabinetU1 | 1 | Distribution, metering and lithium UPS ride-through |
| Coolant distribution unit | 1 | Liquid-to-liquid exchanger, duty and standby pumps, filtration |
| Controls panel | 1 | Building management, access control and monitoring |
| Clean-agent fire suppression | 2 | Cylinders with detection, beside the entrance |
| Heat-rejection packageT1 | 1 | Dry cooler with open top, outside the module |
Options.
Chosen at quotation and fixed in the bill of materials for the edition ordered. Prices are given by quotation.
Compute
- B300 edition: four HGX B300 nodes
- Vera Rubin edition: four HGX Vera Rubin NVL8 nodes
- Two-node start, plant sized for four
Storage
- 0.25, 0.5 or 1.0 PB usable NVMe
- Immutable backup tier
Network
- InfiniBand or Spectrum-X Ethernet scale-out
- Dual-homed uplinks to the customer network
- Out-of-band management network
Power and autonomy
- Lithium UPS ride-through
- Single or dual utility feed
- Generator interface in the graphite end
Cooling and climate
- Dry cooler sized for the site ambient
- Adiabatic assist only where water is available
- Package in louvre fascia or sleeved finish
Security
- Access control and CCTV
- Customer-held encryption keys
- Telemetry under customer control
Site and services
- Factory, site and workload acceptance
- Remote monitoring
- In-country support and spares
Finish
- Pearl white sleeve, the reference finish
- Graphite sleeve with a white logo
- Titanium-grey sleeve for coastal and industrial sites
Best fit.
Where Core Cell C32 earns its place.
- Enterprise private AI cloud
- Banking, insurance, healthcare and industrial analytics
- Production inference, fine-tuning and batch training
- In-country data residency and operational control
Price Core Cell C32 for your site.
Tell us the quantity, the delivery country and the timing. We reply with a written quotation: price, lead time and delivery terms.