TONGAI
QuadBox liquid-cooled compute node

Tongai Technology · QuadBox

80kW of liquid-cooled compute, anywhere.

A self-contained, IP67-rated compute node delivering four times the density of an air-cooled rack. Deploy on-premise or consume as a cloud GPU-as-a-Service — same node, same efficiency, two commercial models.

80kWPower per unit
<1.05Power usage effectiveness
98%Waste heat recovered
Density vs air cooling

Thermal Alpha™ Technology

Direct liquid cooling, engineered for net-positive infrastructure.

Thermal Alpha™ removes heat at the chip with a closed-loop dielectric fluid, eliminating the fans, raised floors, and CRAC units that dominate air-cooled TCO. The recovered heat — 98% of input energy — is redirected to district heating, desalination, or greenhouse agriculture, turning a cost centre into a circular revenue stream.

  • Closed-loop dielectric fluid — no evaporation, no water draw
  • Chip-level heat removal — no hot-aisle containment needed
  • Plug-and-play into any Tongai data centre stand
  • Heat reuse partnerships with municipal and agricultural offtakers
QuadBox internal liquid-cooling assembly

Two ways to deploy

Premise-based. Or cloud. Same node.

QuadBox is one product with two commercial models — own it on your site, or consume it as a metered cloud service from a Tongai campus.

Premise-based

QuadBox On-Premise

Own and operate QuadBox nodes inside your facility or a Tongai data centre. Full capital control, on-site heat reuse, and zero dependency on a third-party cloud.

  • CapEx ownership with 25-year asset life
  • Direct liquid cooling — no raised-floor CRAC retrofit
  • Recovered heat for district heating, desalination, or greenhouses
  • Managed-service SLA option via Tongai Technology
Cloud service

QuadBox GPU-as-a-Service

Consume QuadBox capacity on-demand from a Tongai-operated campus. Per-hour GPU billing, no capital outlay, and the same liquid-cooled efficiency as the on-premise node.

  • Per-hour H100 / A100 billing
  • Burst from 1 GPU to full 80kW node
  • Tongai Connect OPGW backhaul included
  • Carbon-attributable compute — PUE < 1.05

Model comparison

Which QuadBox model fits your workload?

Attribute
On-Premise
GPU-as-a-Service
All-in TCO
Lowest in class
Lowest, no CapEx
Time to compute
6–9 mo deployment
Hours
Heat recovery
100% on-site reuse
Shared campus reuse
Capital model
CapEx + managed SLA
OpEx per GPU-hour
Best for
Sustained 24/7 workloads
Elastic / bursty workloads

Specifications

QuadBox datasheet

Power density

80 kW / unit

PUE

< 1.05

Heat recovery

98%

Cooling

Thermal Alpha™ DLC

GPUs

NVIDIA H100 / A100

Form factor

IP67 self-contained

Deployment

Premise or cloud

Heat reuse

District heating / desalination

Ready to deploy QuadBox?