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

info@worldoneindia.netWhatsApp +91 93244 08321

Modular and factory scale

Edge Cell E32

Thirty-two GPUs at the edge.

Four Rack R8 systems in one ten-foot module, air-cooled, for regional inference, computer vision and visualisation close to where the data is made. It arrives on one truck and needs no outside plant.

Illustrative exterior rendering of Edge Cell E32, shown in full.
32RTX PRO 6000 GPUs
128,000AI TOPS, FP4 sparse
3.07 TBGDDR7 in total
25 to 40 kWIT design envelope, air-cooled

StatusBudgetary architecture. Mechanical and thermal release pending.

Questions first? Ask the team about Edge Cell E32

A closer look.

One ten-foot module: compute, cooling, power and fire protection inside the sleeve.

At the quayInference where the data is made: cameras, cranes and sensors at a container terminal, served from a module that arrives on one truck.
Four racks, one aisleRoof off: four Rack R8 systems with 32 RTX PRO 6000 GPUs along one wall, a front aisle, and the cooling and power held in the graphite ends.
Inside the edge cellFour Rack R8 systems in one row, each with an eight-GPU server at mid height; the power cabinet in the corner and two precision air units on the end wall.

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

Two configurations.

The RTX PRO 6000 reference, and a proposed B300 option in the same ten-foot module.

SpecificationRTX PRO 6000 edition, E32B300 option, E16 (proposed)
GPUs32 RTX PRO 6000 Blackwell Server Edition16 Blackwell Ultra B300
Systems4 Rack R8, one eight-GPU server each2 HGX B300 eight-GPU nodes
Host CPUs4 dual-socket x86 hosts, 1 to 2 TB memory each2 dual-socket x86 hosts
FP4 AI compute128,000 AI TOPS sparse; 4,000 per GPU288 PF sparse; 216 PF dense
GPU memory3.07 TB GDDR7 ECC: 96 GB per GPU at 1.6 TB/s4.2 TB HBM3e: 2.1 TB per node
GPU interconnectPCIe; 32 independent GPUsNVLink 5 in each node: 1.8 TB/s per GPU
Network100, 200 or 400 GbE serving fabric1.6 TB/s scale-out per node
Storage100 to 500 TB usable NVMe100 to 500 TB usable NVMe
IT power25 to 40 kW design envelopeAbout 30 kW, planning allowance
CoolingAir: two precision wall unitsAir-cooled nodes; thermal validation required
StatusBudgetary architecture; engineering release pendingProposed; outside Portfolio 2026

There is no Vera Rubin Edge Cell: the Rubin platforms are liquid-cooled and the ten-foot body carries no coolant loop. Aggregate GPU memory is the sum of independent GPUs, not one pool. The B300 option is proposed and subject to thermal validation. 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.

Nothing outside the module: cooling, power, fire and access are all inside the sleeve.

Modules and packages

1 Ten-foot module
2.99 × 2.44 × 2.90 m, ISO footprint, high-cube height

What the site provides

  • Low-voltage supply sized to the IT envelope and the air units
  • Fibre uplink to the customer network
  • No water: the cell is air-cooled
  • Level concrete pad or steel skid, truck and crane access
ItemQtyWhat it is
Rack R8 systemsR1 to R44Each: one eight-GPU server with a dual-socket x86 host, 1 to 2 TB memory, 60 to 120 TB NVMe and a top-of-rack switch
RTX PRO 6000 Blackwell Server Edition3296 GB GDDR7 ECC each, passively cooled, eight per Rack R8
Precision air units2Wall-mounted on the far end wall; condensers in the graphite end
Power cabinet1Distribution, metering and lithium UPS ride-through
Clean-agent fire suppression1Cylinder with detection, beside the entrance
Overhead cable tray1Serving network and out-of-band management
Access control1At the entrance in the near graphite end

Options.

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

Compute

  • E32: four Rack R8 with 32 RTX PRO 6000 Server Edition, the reference
  • E16: two HGX B300 nodes, proposed
  • Two or three Rack R8 for a phased start, plant sized for four

Storage

  • 100, 250 or 500 TB usable NVMe
  • Local replication to a second cell

Network

  • 100, 200 or 400 GbE serving fabric
  • Single or dual uplinks to the customer network
  • Out-of-band management network

Power and autonomy

  • Lithium UPS ride-through sized to the site's generator start
  • Single or dual utility feed
  • Generator connection in the graphite end

Cooling and climate

  • Air units sized to the design ambient of the site
  • Hot-climate or cold-climate package
  • Intake and condenser clearances confirmed on site

Security

  • Access control at the entrance
  • CCTV and intrusion detection
  • Telemetry under customer control

Site and services

  • Concrete pad or steel skid
  • Factory acceptance before shipping, site acceptance on arrival
  • 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 Edge Cell E32 earns its place.

  1. Telecom and industrial edge inference
  2. Regional AI services and virtual workstations
  3. Computer vision at ports, airports and factories
  4. Digital twins and multi-user visualisation close to the data

Price Edge Cell E32 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.
Edge Cell E3232 GPUs · Modular regional inference cell
Request a price