An AI data centre hosts accelerated compute clusters that process intensive machine learning training, fine-tuning and inference. These workloads rely on graphics processing units (GPUs) and specialised accelerators. Not all AI requires liquid cooling; suitability depends on the hardware heat profile and facility power and cooling design, with liquid cooling serving as an option for higher-density configurations.
Technical suitability depends on the split between gross electrical load and usable IT load, supported liquid cooling topologies, and substation energisation schedules. Assessing facilities across Australian colocation and data centres in Australia requires auditing hall-level mechanical delivery rather than headline campus ratings. Sourcing decisions also connect with frameworks for sovereign AI in Australia and wider power factors covered in our Australian data centre overview.
Rack density and liquid cooling specifications
Facility certification realities
Marketing collateral often claims AI readiness without standard proof. In Macquarie Data Centres' published view, there's currently no single, recognised certification that proves an entire data centre is AI-ready (Macquarie Data Centres). Capacity buyers must inspect hall-level mechanical delivery directly.
To evaluate liquid-cooled colocation capacity across Australian operators, submit your deployment brief through Data Centre Axis capacity sourcing.
AI data centre locations in Sydney and Melbourne
Sydney operational facilities
In Sydney, NEXTDC confirms that its S6 facility in Artarmon is operational, offering customised infrastructure to support AI and HPC workloads (NEXTDC). The facility provides a target IT capacity of 13.5 MW (NEXTDC) and supports densities up to 130 kW per rack maximum (NEXTDC). This target IT capacity represents planned design capacity rather than delivered customer load or vacant space. At Equinix SY5 in Alexandria, SouthernCrossAI (SCX) operates an operational compute node (SouthernCrossAI (SCX)). This node reflects an active customer deployment rather than whole-campus AI certification or general colocation vacancy. Track local options in Sydney data centres.
| Operator | Facility | Location | Verified Operational Scope | Verified Density & IT Capacity |
|---|---|---|---|---|
| NEXTDC | S6 Sydney | Artarmon, NSW | Operational infrastructure supporting AI and HPC workloads (NEXTDC) | 13.5 MW target IT capacity (NEXTDC); up to 130 kW/rack maximum (NEXTDC) |
| Equinix | SY5 | Alexandria, NSW | Operational compute node deployed by SouthernCrossAI (SCX) |
Melbourne procurement diligence
When evaluating facilities across Melbourne data centres, buyers should request hall-specific and rack-specific evidence for power delivery and cooling capacity rather than relying on campus-level claims. Technical teams must directly inspect mechanical plant specifications, liquid cooling loop provisioning, and energised IT power allocations within the nominated data hall.
High-density procurement comparison
When comparing colocation facilities for accelerated computing deployments, buyers should evaluate specific technical parameters across prospective operators:
-
What specific IT capacity is allocated versus gross facility power? Clarify whether quoted megawatt figures represent dedicated, uninterruptible IT capacity provisioned for tenant equipment, or total site gross power incorporating central cooling plant and building infrastructure.
-
What maximum power density is supported per rack position? Confirm whether the server hall sustains target densities, across contiguous footprints without requiring unpopulated adjacent cabinets or thermal derating.
-
Which cooling distribution topologies are installed and operational? Establish whether the hall provides operational direct-to-chip liquid loops, rear-door heat exchangers, or perimeter air delivery, including verified facility supply water temperatures and heat rejection limits.
-
What is the energisation and fit-out timeline for contracted capacity? Determine whether required electrical capacity is energised at the hall power distribution units today, or whether deployment depends upon phased substation commissioning and future fit-out schedules.
-
What structural and network connectivity specifications are delivered? Verify floor loading limits for high-density GPU racks and confirm on-site cross-connect options to interconnection fabrics for distributed computing clusters.
Power allocation, gross electrical load and federal expectations
Gross capacity versus live IT load
Rated grid capacity represents the maximum grid-connected load available to a facility, covering IT load alongside mechanical cooling, power distribution, lighting and auxiliary services (AEMO / Oxford Economics). Mechanical cooling, transformers, uninterruptible power supply (UPS) systems and auxiliary equipment consume incoming power, meaning gross utility connection capacity does not equal usable compute power.
Staged energisation schedules
Facilities typically ramp up over several years as server halls are fitted out, tenants progressively utilise contracted space and computing workloads increase (AEMO / Oxford Economics). Buyers must verify contracted delivery dates for each power increment. Further connection context is detailed in our Australian data centre overview.
Federal data centre expectations
-
Grid cost sharing: The Energy and Climate Change Ministerial Council agreed rule change requests be progressed to ensure data centre developers cover their share of transmission and distribution infrastructure costs (Australian Energy Market Commission (AEMC)).
-
Regulatory status: Consultations on rule change requests ERC0448 and ERC0456 opened in 2026 through the AEMC; binding National Electricity Rules (NER) amendments remain in consultation and are not yet enacted statutory law.
Sourcing AI data centre capacity in Australia
Technical validation at the rack position
Procurement requires auditing electrical feeds and heat dissipation at target racks. Capacity buyers should require the provider to demonstrate how power, cooling, space and operating models support high-density hardware at the proposed rack position. Compare options in colocation in Australia.
Cooling distribution verification
For liquid-cooled hardware, buyers must confirm mechanical connectivity:
-
Require the provider to demonstrate how liquid cooling loops connect to facility infrastructure and confirm where the coolant distribution unit (CDU) will sit.
-
Inspect secondary pipework headers, isolation valves, and commissioning schedules at row manifolds.
Colocation, wholesale and AI-ready capacity requirements are matched to facilities manually, followed by direct introductions to operators where both parties agree.
Our team conducts an initial review of your requirement and commercial evidence. Following agreed sharing permissions, we facilitate targeted introductions between principals. Where required, Data Centre Axis delivers individually scoped advisory support and introduces specialist consultants for technical, planning and legal requirements.
AI data centre questions
What is an AI-ready data centre?
An AI-ready data centre supports accelerated compute with high-density power distribution, liquid cooling options where thermal density requires it. In Macquarie Data Centres' published view, there's currently no single, recognised certification that proves an entire data centre is AI-ready (Macquarie Data Centres).
Where are AI data centres located in Melbourne and Sydney?
In Sydney, verified deployments include NEXTDC S6 in Artarmon (NEXTDC) and the SouthernCrossAI operational compute node at Equinix SY5 in Alexandria (SouthernCrossAI (SCX)). In Melbourne, buyers should check the hall-specific supported power and cooling profile directly with operators. Compare local facilities across Sydney data centres and Melbourne data centres.
What does an AI data centre actually do?
An AI data centre houses clusters of graphics processing units (GPUs) and specialised accelerators executing machine learning training, fine-tuning and inference workloads. Not all AI requires liquid cooling; suitability depends on the hardware heat profile and facility power and cooling design, with liquid cooling offered as an option for higher-density thermal loads.
Does advertised campus megawatt capacity equal available AI compute power?
No. Advertised megawatt figures can represent target IT capacity, staged phase allocations, or gross grid connections that include facility cooling and building services. Capacity buyers must verify whether stated figures denote usable IT load or gross site power. Furthermore, facilities typically ramp up over several years as server halls are fitted out, tenants progressively utilise contracted space and computing workloads increase (AEMO / Oxford Economics).
Assess high-density AI capacity options and confirm liquid cooling commissioning schedules through Data Centre Axis capacity sourcing.