Data centre networking covers the switching, routing and optical systems that move data within a data hall, between halls and campuses, and out to carriers and clouds. AI has changed the category. GPU clusters, such as the NVIDIA GB300 deployments at Firmus's Project Southgate sites, need back-end fabrics that connect thousands of accelerators with very high bandwidth and low latency, on top of the conventional front-end network.
For physical fibre routes, subsea access and dark fibre, see data centre fibre.
The Role of Networking in Data Centres
| Layer | What it does |
|---|---|
| Front-end network | Connects servers to storage, users and the internet: leaf-spine switching and routing |
| Back-end (AI) fabric | Connects GPUs for distributed training, using very high-bandwidth, low-latency links |
| Storage network | High-throughput links to storage systems |
| Data centre interconnect (DCI) | Optical links between halls, buildings and campuses |
| Edge and peering | Routers connecting to carriers, internet exchanges and cloud on-ramps |
| Out-of-band management | Separate network for managing equipment during failures |
Key Infrastructure Requirements
Switching and routing
Leaf-spine architectures give predictable latency and scale. Operators and tenants choose platforms, and integrators design, deploy and support them.
Optical networking
Coherent optics and wavelength systems for DCI between buildings and campuses. Campuses spread across Western Sydney need DCI back to inner-city interconnection points.
Interconnection
Cross-connects in meet-me rooms, internet exchange ports, and cloud on-ramps. Interconnection-focused operators (Equinix, Global Switch, NEXTDC) sell this as a core service.
High bandwidth
AI clusters need far more bandwidth per rack than conventional halls, which means dense optical cabling, many transceivers and careful cable management in overhead pathways.
AI clusters
GPU back-end fabrics are closely tied to the server platform, and their design affects rack layout, cable lengths and power. They are usually delivered by the tenant or GPU cloud provider with specialist integrators, inside space provided by the data centre operator.
Australian Networking Suppliers
Directory listings for this category are being verified before they are published. Supply to data centre projects? Add your company to be considered.
| Type | Typical scope |
|---|---|
| Network equipment vendors and distributors | Switches, routers, optics |
| Systems integrators | Design, deployment, configuration, support |
| Optical transport specialists | DCI, wavelength systems |
| Structured cabling contractors | High-density fibre cabling, testing |
| Interconnection providers | Exchanges, cross-connect services |
Selecting a Supplier
- Experience with the tenant's chosen platforms, and with GPU fabrics where relevant
- Staff certified on the vendor platforms and cabling standards being deployed
- Deployment capacity: installing and testing thousands of links on tight fit-out programmes
- Security clearances for government facilities (see Canberra)
- Local support and spares
Current Project Demand
- AI deployments: Firmus Southgate (Tasmania and Melbourne), NEXTDC S6 and future M4 and S7 AI halls, Macquarie IC3 Super West
- Cloud expansion: AWS (A$20 billion, 2025–2029) and Microsoft (A$25 billion to 2029)
- New campuses needing DCI back to interconnection hubs
Relevant Data Centre Projects
AI data centres in Australia · All projects
List Your Company
Grow Your Data Centre Pipeline
AISB helps networking integrators and vendors reach the operators, GPU cloud providers and enterprises building Australian AI and cloud capacity.
See something that needs updating? Submit a correction.
