Australia's resilience problem isn't redundancy. It's shared infrastructure

Ask most operators if their network is resilient and you'll hear a familiar response: dual providers, separate contracts, and a failover path engineered into the design. It satisfies the checklist and looks reassuring on paper. The problem is that true resilience isn't measured by architecture diagrams, it's more so tested when things go wrong.

The industry's definition of resilience has become overly reliant on architecture diagrams and procurement checklists. Yet networks do not fail on paper; they fail in the real world.

When a fibre duct is damaged, the disruption affects every service carried within it, regardless of how many providers are represented on the contract. If primary and backup connections share the same physical route, they share the same vulnerability.

This exposes a critical misunderstanding in the market and where the industry continues to blur an important distinction. Redundancy and resilience are not the same thing.

Redundancy provides an alternative path when something fails. Resilience requires that a single failure cannot affect both paths in the first place. For years, many organisations have assumed they were buying resilience when they were actually buying redundancy. The difference is often invisible until the moment it matters most.

'Separate providers' doesn't always mean separate infrastructure

The reason this resilience gap persists is rooted in the way Australia's fibre networks were originally built.

Much of the country's fibre infrastructure traces back to a relatively small number of foundational networks, constructed along common routes and through shared conduit. Over time, those assets have been sold, leased, rebranded and repackaged by different providers. While the commercial ownership may have evolved, the underlying physical infrastructure often has not.

As a result, organisations can purchase services from multiple providers believing they have achieved network diversity, when in reality those services may still share the same physical path for part of their journey. The contracts are different. The brands are different. But the fibre itself may be running through the same conduit, separated by little more than a few centimetres.

This creates a dangerous illusion of resilience. Supplier diversity is not the same as infrastructure diversity. When disruption occurs, it is physical risk and not commercial arrangements that determines whether services stay online.

AI is changing the economics of network failure

Historically, this was a risk many organisations were willing to accept. Outages were inconvenient, but most systems could tolerate a temporary interruption.

That equation is changing rapidly.

Modern digital environments are more distributed, interconnected and dependent on continuous connectivity than ever before. A failure between two facilities no longer impacts a single application or process. It can trigger a cascading effect across every service that relies on that connection, with consequences extending far beyond the duration of the initial outage.

Artificial intelligence is amplifying those requirements. AI workloads operate at a scale and intensity where interruptions are not simply disruptive but incredibly costly. Even brief periods of lost connectivity can create downstream impacts that persist long after services are restored.

Equally important, AI infrastructure derives value from connectivity itself. Data centres, cloud environments and compute clusters only function as intended when they can communicate seamlessly with one another. A network interruption can therefore compromise the effectiveness of an entire environment, even when every server remains operational.

In this context, resilience is no longer a desirable feature but a prerequisite. 

What true path diversity requires

True resilience is a physical characteristic, not a contractual one. It requires genuine separation at the infrastructure layer: distinct duct routes, independent physical paths, and no shared points of failure between primary and backup services.

That is a far higher standard than many organisations realise. When network routes are examined in detail, assumptions about diversity often prove inaccurate.

It is also why resilience must be designed into infrastructure from the outset rather than added later. Physical diversity cannot be retrofitted through contracts, service level agreements or additional providers. It must be engineered into the network itself.

For FibreconX, that principle informed every stage of network design. From operating independent duct infrastructure and creating physically diverse routes between key data centre locations, to ensuring sufficient capacity so congestion does not become a resilience risk in its own right.

The broader point extends beyond any individual network. Physical diversity is a deliberate design choice. If it is not built in from the beginning, it is exceptionally difficult to create later.

The questions organisations should be asking

For organisations supporting critical workloads, think AI inference, large scale training environments, hyperscale connectivity or other mission critical operations, the traditional questions around redundancy are no longer sufficient.

Instead, leaders should be asking three simple questions:

• Do we know the physical routes our primary and backup services actually take?
• Has our provider confirmed that those routes share no common conduit or physical infrastructure?
• Is our resilience strategy based on verified engineering, or on assumptions?

The answers are often more revealing than any service level agreement.

An SLA defines what happens after an outage occurs. Physical path diversity determines whether the outage happens at all.

And in critical infrastructure, the failures that matter most are rarely the ones organisations anticipate. They are the ones hidden beneath the assumptions nobody thought to challenge.a