Vertiv
Liebert XDU — sovereign cloud + colocation deployments in Sydney, Melbourne. ASHRAE W32 / W40 alignment.
Data Centre Liquid Cooling · Phase 1 Primary
Direct-to-chip liquid cooling cold plate kits for Australian colocation, sovereign cloud, and HPC refresh deployments. CDU-vendor compatible with Vertiv, CoolIT, Stulz, and Motivair. Rack power density 80–120 kW+ supported — the step-change from air-cooled 8–25 kW that enables high-density GPU clusters and AI training infrastructure. Cold plates designed, engineered & manufactured in Huizhou; AS/NZS & ChAFTA documentation from Carlton VIC.
Solution Snapshot · 4 KPI
Engineering Reference & Standards
AU colocation operators deploying liquid-cooled GB200 NVL72 racks for AI tenants. Pre-assembled cold plate + manifold + Stäubli RBE03 quick-disconnect kits, helium-leak verified.
Australian sovereign cloud builds for defence, intelligence, and critical-infrastructure workloads. Carlton-coordinated engineering response in AEST/AEDT business hours.
Australian HPC supercomputing refresh windows. Direct-to-chip cooling 80–120 kW per rack, AS/NZS 60068-2-14 thermal cycling validated, lot-traceable per shipment.
Section 1 of 3
| Cooling architecture | Rack density | Practical limit | Suitable for |
|---|---|---|---|
| Air-cooled (hot/cold aisle) | 8–15 kW | Hot-aisle / cold-aisle airflow constraint | Legacy compute / general-purpose servers |
| Air-cooled (rear-door heat exchanger) | 15–25 kW | Rear-door coil flow rate, supplemental water | Mid-density compute |
| Direct-to-chip liquid (entry) | 25–60 kW | Cold plate ΔP budget, CDU pump curve | Mixed GPU / CPU clusters, HPC entry |
| Direct-to-chip liquid (high-density) | 80–120 kW+ | CDU heat-rejection capacity to facility loop | NVIDIA GB200 NVL72, AI training clusters |
| Immersion (single-phase / two-phase) | 100–200 kW+ | Coolant compatibility / capex | Specialty HPC / research compute |
Section 2 of 3
Liebert XDU — sovereign cloud + colocation deployments in Sydney, Melbourne. ASHRAE W32 / W40 alignment.
CHx series — Pawsey, NCI HPC refresh and sovereign-cloud private regions.
/ 200 — mid-tier AU integrators and regional colocation operators.
— HPC and AI training cluster builds. ASHRAE W40 deployment.
Section 3 of 3
Microsoft Azure AU East / AU Southeast, AWS Sydney, Oracle / SAP private regions deployed across Sydney, Melbourne, Perth.
Pawsey Supercomputing Research Centre (Perth), NCI Gadi (Canberra), AARNet research-network sites.
NEXTDC, Equinix AU, Macquarie Telecom, AirTrunk — high-density rack programmes for AI tenants.
Optus, Telstra, TPG edge AI deployments + 5G fronthaul thermal hardware.
Civilian and commercial government integrator builds within engagement scope.
Server OEMs and rack-builder integrators serving the AU data centre channel.
80–120 kW+ per rack vs 8–25 kW for air cooling. Total rack-level ΔP envelope 25–45 kPa across cold plate + manifold + interconnects + CDU heat exchanger.
Vertiv (Liebert XDU), CoolIT (CHx series), Stulz CMU, Motivair. Cold plate hydraulic envelope tuned during DFM iteration to align with your CDU pump curve and chemistry.
Sovereign cloud (Azure / AWS / Oracle), HPC refresh (Pawsey, NCI), AARNet, mid-tier colocation (NEXTDC, Equinix AU, Macquarie, AirTrunk), edge AI & telecom, civilian government integrators.
Continue Exploring
When AI data centres need direct-to-chip liquid cooling over air.
SOLUTIONDirect-to-chip cold plate kits for NVIDIA GB200, GB300, H200, AMD SP5, Intel Birch Stream — primary AU AI server platforms.
View solution → PRODUCTGB200 / GB300 / H200 kits for Australian sovereign cloud and HPC refresh deployments.
View product → CAPABILITYDocumentation handoff for AS/NZS 5139, 4777, 60068-2-14, 60529 + ChAFTA per shipment.
View capability → RESOURCE12-section engineering reference for direct-to-chip AI server cold plate design.
Read the guide →TDP map per platform, target ΔP and flow rate, CDU model, qualification scope. Carlton-coordinated 24-hour response with technical review by Huizhou.
Request a QuoteData centre liquid cooling by city