Request a quoteBuy GPUs
HYPERCUBE
HYPERQ
GPUs
Astra Cybercity
Sectors
Press Room
Company

info@worldoneindia.netWhatsApp +91 93244 08321

Modular and factory scale

Factory 9 G576

The SuperPod, in nine modules.

Eight compute modules and one control module, built and tested in the factory as one block, with two cooling bars and a power bar. Eight racks is also the unit of NVIDIA's Vera Rubin SuperPOD, so the block carries into the next generation.

Illustrative exterior rendering of Factory 9 G576, shown in full.
576GPUs in eight NVL72 racks
11.52 EFFP4 sparse; 28.8 EF NVFP4 in the Vera Rubin edition
296 TBfast memory, GB300 edition
9factory-built modules: eight compute, one control

StatusReference architecture; supplier quote required. Vera Rubin edition: next generation, allocation and certification.

Questions first? Ask the team about Factory 9 G576

A closer look.

Nine factory-built modules on one pad, with two cooling bars and a power bar.

Desert factoryNine modules in open desert. Dry coolers, no evaporative water, and a footprint that ships on trucks.
The factory, openedFour roofs lifted: one NVL72 rack, its power cabinet and its coolant unit in each compute module; the fabric and storage racks in the control module.
A compute moduleOne of eight: a single NVL72 rack between its power cabinet and its coolant unit, with the pipes running on through the end wall to the cooling bar.
The control moduleThree fabric racks by the entrance and four storage racks beyond: the scale-out fabric, management and storage that serve all eight compute modules.

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.

SpecificationGB300 editionVera Rubin edition
Racks8 GB300 NVL728 Vera Rubin NVL72
GPUs576 Blackwell Ultra B300576 Rubin
CPUs288 Grace: 20,736 cores288 Vera: 25,344 cores
FP4 inference11.52 EF sparse28.8 EF NVFP4
FP4 dense or training8.64 EF dense20 EF NVFP4 training
FP85.76 EFAbout 10 EF, derived
Fast memory296 TB: 160 TB HBM3e, 136 TB LPDDR5XAbout 598 TB: 166 TB HBM4, 432 TB LPDDR5X
NVLink domains8 × 130 TB/s8 × 260 TB/s
Scale-out800G cross-rack fabric1.6 Tb/s per GPU
Storage2 to 5 PB class2 to 5 PB class
IT loadAbout 1.25 to 1.45 MWAbout 1.7 to 2.1 MW, planning
Facility1.45 to 1.75 MW at PUE 1.15 to 1.20About 2.0 to 2.5 MW, planning
Delivery16 to 20 weeks after procurement, compliance and site gatesPer NVIDIA allocation
StatusReference architecture; supplier quote requiredNext generation; allocation and certification

Vera Rubin IT and facility ranges are first-principles planning ranges built from 190 to 230 kW racks plus the control module and auxiliaries. External power and heat rejection are separately defined site-interface packages. 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.

Nine factory-built modules, two cooling bars and one power bar on a single pad. Site power and water are site-interface packages.

Modules and packages

8 Compute modules C1 to C8
Twenty-foot, 6.06 × 2.44 × 2.90 m each
1 Control module S1
Twenty-foot, set crosswise at the head of the street
2 Cooling bars T1 and T2
Dry coolers, 27.2 m long, along the outer sides
1 Power bar U1
Switchgear and UPS at the open end of the street

What the site provides

  • Medium-voltage supply to the power bar
  • Diverse fibre routes into the control module
  • No water in operation: closed loops to the dry-cooler bars
  • One pad for nine modules, two bars and the power bar
ItemQtyWhat it is
Compute modulesC1 to C88Each: one NVL72 rack, one coolant distribution unit, one power cabinet, one fire cylinder
NVL72 racks8GB300 NVL72, or Vera Rubin NVL72 in the Vera Rubin edition
Coolant distribution units8One per compute module, each on its own loop
Power cabinets8One distribution and protection cabinet per compute module
Control moduleS11Three fabric racks and four storage racks
Fabric racks3800G scale-out fabric, management and out-of-band
Storage racks42 to 5 PB class NVMe
Cooling barsT1, T22Dry coolers with open tops; site-interface package
Power barU11MV and LV switchgear and UPS; site-interface package

Options.

Chosen at quotation and fixed in the bill of materials for the edition ordered. Prices are given by quotation.

Compute

  • GB300 edition: eight GB300 NVL72
  • Vera Rubin edition: eight Vera Rubin NVL72
  • Phased build: four compute modules first

Storage

  • 2 PB, 3.5 PB or 5 PB class
  • Performance and context tiers

Network

  • InfiniBand or Spectrum-X scale-out
  • Multi-tenant segmentation
  • Out-of-band management network

Power and autonomy

  • 2N or N+1 power bar
  • UPS ride-through sized to generator start
  • 800 VDC option with Vera Rubin

Cooling and climate

  • Dry-cooler bars sized for the design ambient
  • Hot-climate bars
  • Bars in louvre fascia or sleeved finish

Security

  • Site perimeter and access control
  • Customer-held keys
  • Telemetry under customer control

Site and services

  • Factory acceptance of the whole block
  • Site and workload acceptance
  • In-country operations 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 Factory 9 G576 earns its place.

  1. National AI capacity and sovereign multi-tenant research
  2. Telecom and GPU-as-a-service platforms
  3. Frontier-model inference and fine-tuning
  4. Regulated workloads under in-country control of the full stack

Price Factory 9 G576 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.

Developed by Quantum AI Connect, owner of HYPERCUBE and HYPERQ.
Factory 9 G576576 GPUs · Nine-module AI factory
Request a price