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

info@worldoneindia.netWhatsApp +91 93244 08321

Next generation

Titan 13 G864

SuperPod Ultra, next generation.

Twelve Vera Rubin NVL72 compute modules and one shared control module per block: 864 Rubin GPUs in blocks that tile into a campus. Planned as a phased programme after the SuperPod Ultra, with each block accepted on its own.

Illustrative exterior rendering of Titan 13 G864, shown in full.
864Rubin GPUs per block
43.2 EFNVFP4 inference per block; 30 EF training
896 TBfast memory: 248 TB HBM4, 648 TB LPDDR5X
13modules per block: twelve compute, one control

StatusNext generation, Vera Rubin only. Planning architecture, quoted once Factory 9 commissioning evidence, Vera Rubin allocation and certification are in place.

Questions first? Ask the team about Titan 13 G864

A closer look.

Thirteen factory-built modules per block, with cooling bars, a power bar and a high-voltage compound.

Campus at duskA campus block beside its substation at dusk. The emerald lines trace each module; the plant stays dark and quiet.
The block, openedFive roofs lifted, across all three rows and on the control module: one NVL72 rack, its power cabinet and its coolant unit in every compute module.
A Vera Rubin compute moduleOne of twelve: a Vera Rubin NVL72 rack between its coolant unit and its power cabinet, seen from the far end towards the entrance.

Styling concepts from the HYPERCUBE Product Family Design Collection, Edition 3. Configuration not finalised; renders do not depict a supplied configuration.

Designed for Vera Rubin.

Figures for one Titan block and for each of its twelve compute modules.

SpecificationTitan 13, Vera RubinPer compute module
Racks12 Vera Rubin NVL721 Vera Rubin NVL72
GPUs864 Rubin72 Rubin
CPUs432 Vera: 38,016 Arm cores36 Vera: 3,168 Arm cores
NVFP4 inference43.2 EF3.6 EF
NVFP4 training30 EF2.5 EF
FP8About 15 EF, derivedAbout 1.26 EF, derived
GPU memoryAbout 248 TB HBM420.7 TB HBM4 at 1.6 PB/s
CPU memory648 TB LPDDR5X54 TB LPDDR5X
NVLink 612 domains × 260 TB/s260 TB/s; 3.6 TB/s per GPU
Scale-outShared fabric in the control module1.6 Tb/s per GPU, ConnectX-9
IT loadAbout 2.6 to 3.2 MW, planningAbout 190 to 230 kW rack
FacilityAbout 3.0 to 3.8 MW, planningOne CDU on 45 °C water
StatusNext generation; planning architectureAllocation and certification

Titan 13 is offered in the Vera Rubin edition only. Power figures are first-principles planning ranges from 190 to 230 kW racks, the control module and auxiliaries, not approved duties. Shared-plant savings are proven before they are credited. 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.

Thirteen factory-built modules, two cooling bars, one power bar and one high-voltage compound on a single campus pad.

Modules and packages

12 Compute modules C1 to C12
Twenty-foot, 6.06 × 2.44 × 2.90 m each, three rows of four
1 Control module S1
Twenty-foot, crosswise at the head of the first street
2 Cooling bars T1 and T2
Dry coolers, 27.2 m long, along the outer rows
2 Power bar U1 and HV compound
Switchgear and UPS in the bar; transformers and switchyard interface in the compound, by site

What the site provides

  • High-voltage intake to the HV compound
  • Diverse fibre routes into the control module
  • No water in operation: closed loops to the dry-cooler bars on 45 °C water
  • One campus pad, crane and transport access to every street
ItemQtyWhat it is
Compute modulesC1 to C1212Each: one Vera Rubin NVL72 rack, one coolant distribution unit, one power cabinet, one fire cylinder
Vera Rubin NVL72 racks12864 Rubin GPUs and 432 Vera CPUs
Coolant distribution units12Sized for 190 to 230 kW racks on 45 °C water
Power cabinets12One per compute module
Control moduleS11Five fabric racks and three storage racks
Fabric racks5Shared scale-out fabric, management and out-of-band
Storage racks3Multi-petabyte class NVMe
Cooling barsT1, T22Dry coolers with open tops; site-interface package
Power barU11Switchgear and UPS; site-interface package
High-voltage compoundHV1Transformers and switchyard interface, by site

Options.

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

Compute

  • Vera Rubin edition only: twelve Vera Rubin NVL72
  • Phased: Factory 9 block first, then the Titan extension

Storage

  • Multi-petabyte, performance and context tiers
  • Immutable backup tier

Network

  • InfiniBand or Spectrum-X scale-out at 1.6 Tb/s per GPU
  • Multi-tenant segmentation
  • Out-of-band management network

Power and autonomy

  • 415 or 480 VAC; 800 VDC option
  • 2N high-voltage intake
  • UPS ride-through sized to generator start

Cooling and climate

  • Dry-cooler bars on 45 °C water
  • Bars sized for the design ambient
  • Louvre fascia or sleeved finish

Security

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

Site and services

  • Separately accepted blocks
  • Factory, 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 Titan 13 G864 earns its place.

  1. National AI programmes and hyperscale sovereign cloud
  2. Power owners monetising large energy positions
  3. Multi-phase campuses built from repeatable, bankable blocks
  4. Research consortia with committed multi-year capacity

Plan Titan 13 for your programme.

Share the intended location, the power available and the phasing. We reply with a planning proposal and the steps to a quotation.

Developed by Quantum AI Connect, owner of HYPERCUBE and HYPERQ.
Titan 13 G864864 Rubin GPUs per block · SuperPod Ultra, next generation
Request a price